1 /*
2 * Copyright 2022 The Android Open Source Project
3 *
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
7 *
8 * http://www.apache.org/licenses/LICENSE-2.0
9 *
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
15 */
16
17 #include <cmath>
18
19 #include "ultrahdr/gainmapmath.h"
20
21 namespace ultrahdr {
22
23 ////////////////////////////////////////////////////////////////////////////////
24 // Framework
25
getReferenceDisplayPeakLuminanceInNits(uhdr_color_transfer_t transfer)26 float getReferenceDisplayPeakLuminanceInNits(uhdr_color_transfer_t transfer) {
27 switch (transfer) {
28 case UHDR_CT_LINEAR:
29 return kPqMaxNits;
30 case UHDR_CT_HLG:
31 return kHlgMaxNits;
32 case UHDR_CT_PQ:
33 return kPqMaxNits;
34 case UHDR_CT_SRGB:
35 return kSdrWhiteNits;
36 case UHDR_CT_UNSPECIFIED:
37 return -1.0f;
38 }
39 return -1.0f;
40 }
41
42 ////////////////////////////////////////////////////////////////////////////////
43 // Use Shepard's method for inverse distance weighting.
44
euclideanDistance(float x1,float x2,float y1,float y2)45 float ShepardsIDW::euclideanDistance(float x1, float x2, float y1, float y2) {
46 return sqrt(((y2 - y1) * (y2 - y1)) + (x2 - x1) * (x2 - x1));
47 }
48
fillShepardsIDW(float * weights,int incR,int incB)49 void ShepardsIDW::fillShepardsIDW(float* weights, int incR, int incB) {
50 for (int y = 0; y < mMapScaleFactor; y++) {
51 for (int x = 0; x < mMapScaleFactor; x++) {
52 float pos_x = ((float)x) / mMapScaleFactor;
53 float pos_y = ((float)y) / mMapScaleFactor;
54 int curr_x = floor(pos_x);
55 int curr_y = floor(pos_y);
56 int next_x = curr_x + incR;
57 int next_y = curr_y + incB;
58 float e1_distance = euclideanDistance(pos_x, curr_x, pos_y, curr_y);
59 int index = y * mMapScaleFactor * 4 + x * 4;
60 if (e1_distance == 0) {
61 weights[index++] = 1.f;
62 weights[index++] = 0.f;
63 weights[index++] = 0.f;
64 weights[index++] = 0.f;
65 } else {
66 float e1_weight = 1.f / e1_distance;
67
68 float e2_distance = euclideanDistance(pos_x, curr_x, pos_y, next_y);
69 float e2_weight = 1.f / e2_distance;
70
71 float e3_distance = euclideanDistance(pos_x, next_x, pos_y, curr_y);
72 float e3_weight = 1.f / e3_distance;
73
74 float e4_distance = euclideanDistance(pos_x, next_x, pos_y, next_y);
75 float e4_weight = 1.f / e4_distance;
76
77 float total_weight = e1_weight + e2_weight + e3_weight + e4_weight;
78
79 weights[index++] = e1_weight / total_weight;
80 weights[index++] = e2_weight / total_weight;
81 weights[index++] = e3_weight / total_weight;
82 weights[index++] = e4_weight / total_weight;
83 }
84 }
85 }
86 }
87
88 ////////////////////////////////////////////////////////////////////////////////
89 // sRGB transformations
90
91 // See IEC 61966-2-1/Amd 1:2003, Equation F.7.
92 static const float kSrgbR = 0.212639f, kSrgbG = 0.715169f, kSrgbB = 0.072192f;
93
srgbLuminance(Color e)94 float srgbLuminance(Color e) { return kSrgbR * e.r + kSrgbG * e.g + kSrgbB * e.b; }
95
96 // See ITU-R BT.709-6, Section 3.
97 // Uses the same coefficients for deriving luma signal as
98 // IEC 61966-2-1/Amd 1:2003 states for luminance, so we reuse the luminance
99 // function above.
100 static const float kSrgbCb = (2 * (1 - kSrgbB)), kSrgbCr = (2 * (1 - kSrgbR));
101
srgbRgbToYuv(Color e_gamma)102 Color srgbRgbToYuv(Color e_gamma) {
103 float y_gamma = srgbLuminance(e_gamma);
104 return {{{y_gamma, (e_gamma.b - y_gamma) / kSrgbCb, (e_gamma.r - y_gamma) / kSrgbCr}}};
105 }
106
107 // See ITU-R BT.709-6, Section 3.
108 // Same derivation to BT.2100's YUV->RGB, below. Similar to srgbRgbToYuv, we
109 // can reuse the luminance coefficients since they are the same.
110 static const float kSrgbGCb = kSrgbB * kSrgbCb / kSrgbG;
111 static const float kSrgbGCr = kSrgbR * kSrgbCr / kSrgbG;
112
srgbYuvToRgb(Color e_gamma)113 Color srgbYuvToRgb(Color e_gamma) {
114 return {{{clampPixelFloat(e_gamma.y + kSrgbCr * e_gamma.v),
115 clampPixelFloat(e_gamma.y - kSrgbGCb * e_gamma.u - kSrgbGCr * e_gamma.v),
116 clampPixelFloat(e_gamma.y + kSrgbCb * e_gamma.u)}}};
117 }
118
119 // See IEC 61966-2-1/Amd 1:2003, Equations F.5 and F.6.
srgbInvOetf(float e_gamma)120 float srgbInvOetf(float e_gamma) {
121 if (e_gamma <= 0.04045f) {
122 return e_gamma / 12.92f;
123 } else {
124 return pow((e_gamma + 0.055f) / 1.055f, 2.4f);
125 }
126 }
127
srgbInvOetf(Color e_gamma)128 Color srgbInvOetf(Color e_gamma) {
129 return {{{srgbInvOetf(e_gamma.r), srgbInvOetf(e_gamma.g), srgbInvOetf(e_gamma.b)}}};
130 }
131
srgbInvOetfLUT(float e_gamma)132 float srgbInvOetfLUT(float e_gamma) {
133 int32_t value = static_cast<int32_t>(e_gamma * (kSrgbInvOETFNumEntries - 1) + 0.5);
134 // TODO() : Remove once conversion modules have appropriate clamping in place
135 value = CLIP3(value, 0, kSrgbInvOETFNumEntries - 1);
136 static LookUpTable kSrgbLut(kSrgbInvOETFNumEntries, static_cast<float (*)(float)>(srgbInvOetf));
137 return kSrgbLut.getTable()[value];
138 }
139
srgbInvOetfLUT(Color e_gamma)140 Color srgbInvOetfLUT(Color e_gamma) {
141 return {{{srgbInvOetfLUT(e_gamma.r), srgbInvOetfLUT(e_gamma.g), srgbInvOetfLUT(e_gamma.b)}}};
142 }
143
144 // See IEC 61966-2-1/Amd 1:2003, Equations F.10 and F.11.
srgbOetf(float e)145 float srgbOetf(float e) {
146 constexpr float kThreshold = 0.0031308f;
147 constexpr float kLowSlope = 12.92f;
148 constexpr float kHighOffset = 0.055f;
149 constexpr float kPowerExponent = 1.0f / 2.4f;
150 if (e <= kThreshold) {
151 return kLowSlope * e;
152 }
153 return (1.0f + kHighOffset) * std::pow(e, kPowerExponent) - kHighOffset;
154 }
155
srgbOetf(Color e)156 Color srgbOetf(Color e) { return {{{srgbOetf(e.r), srgbOetf(e.g), srgbOetf(e.b)}}}; }
157
158 ////////////////////////////////////////////////////////////////////////////////
159 // Display-P3 transformations
160
161 // See SMPTE EG 432-1, Equation G-7.
162 static const float kP3R = 0.2289746f, kP3G = 0.6917385f, kP3B = 0.0792869f;
163
p3Luminance(Color e)164 float p3Luminance(Color e) { return kP3R * e.r + kP3G * e.g + kP3B * e.b; }
165
166 // See ITU-R BT.601-7, Sections 2.5.1 and 2.5.2.
167 // Unfortunately, calculation of luma signal differs from calculation of
168 // luminance for Display-P3, so we can't reuse p3Luminance here.
169 static const float kP3YR = 0.299f, kP3YG = 0.587f, kP3YB = 0.114f;
170 static const float kP3Cb = 1.772f, kP3Cr = 1.402f;
171
p3RgbToYuv(Color e_gamma)172 Color p3RgbToYuv(Color e_gamma) {
173 float y_gamma = kP3YR * e_gamma.r + kP3YG * e_gamma.g + kP3YB * e_gamma.b;
174 return {{{y_gamma, (e_gamma.b - y_gamma) / kP3Cb, (e_gamma.r - y_gamma) / kP3Cr}}};
175 }
176
177 // See ITU-R BT.601-7, Sections 2.5.1 and 2.5.2.
178 // Same derivation to BT.2100's YUV->RGB, below. Similar to p3RgbToYuv, we must
179 // use luma signal coefficients rather than the luminance coefficients.
180 static const float kP3GCb = kP3YB * kP3Cb / kP3YG;
181 static const float kP3GCr = kP3YR * kP3Cr / kP3YG;
182
p3YuvToRgb(Color e_gamma)183 Color p3YuvToRgb(Color e_gamma) {
184 return {{{clampPixelFloat(e_gamma.y + kP3Cr * e_gamma.v),
185 clampPixelFloat(e_gamma.y - kP3GCb * e_gamma.u - kP3GCr * e_gamma.v),
186 clampPixelFloat(e_gamma.y + kP3Cb * e_gamma.u)}}};
187 }
188
189 ////////////////////////////////////////////////////////////////////////////////
190 // BT.2100 transformations - according to ITU-R BT.2100-2
191
192 // See ITU-R BT.2100-2, Table 5, HLG Reference OOTF
193 static const float kBt2100R = 0.2627f, kBt2100G = 0.677998f, kBt2100B = 0.059302f;
194
bt2100Luminance(Color e)195 float bt2100Luminance(Color e) { return kBt2100R * e.r + kBt2100G * e.g + kBt2100B * e.b; }
196
197 // See ITU-R BT.2100-2, Table 6, Derivation of colour difference signals.
198 // BT.2100 uses the same coefficients for calculating luma signal and luminance,
199 // so we reuse the luminance function here.
200 static const float kBt2100Cb = (2 * (1 - kBt2100B)), kBt2100Cr = (2 * (1 - kBt2100R));
201
bt2100RgbToYuv(Color e_gamma)202 Color bt2100RgbToYuv(Color e_gamma) {
203 float y_gamma = bt2100Luminance(e_gamma);
204 return {{{y_gamma, (e_gamma.b - y_gamma) / kBt2100Cb, (e_gamma.r - y_gamma) / kBt2100Cr}}};
205 }
206
207 // See ITU-R BT.2100-2, Table 6, Derivation of colour difference signals.
208 //
209 // Similar to bt2100RgbToYuv above, we can reuse the luminance coefficients.
210 //
211 // Derived by inversing bt2100RgbToYuv. The derivation for R and B are pretty
212 // straight forward; we just invert the formulas for U and V above. But deriving
213 // the formula for G is a bit more complicated:
214 //
215 // Start with equation for luminance:
216 // Y = kBt2100R * R + kBt2100G * G + kBt2100B * B
217 // Solve for G:
218 // G = (Y - kBt2100R * R - kBt2100B * B) / kBt2100B
219 // Substitute equations for R and B in terms YUV:
220 // G = (Y - kBt2100R * (Y + kBt2100Cr * V) - kBt2100B * (Y + kBt2100Cb * U)) / kBt2100B
221 // Simplify:
222 // G = Y * ((1 - kBt2100R - kBt2100B) / kBt2100G)
223 // + U * (kBt2100B * kBt2100Cb / kBt2100G)
224 // + V * (kBt2100R * kBt2100Cr / kBt2100G)
225 //
226 // We then get the following coeficients for calculating G from YUV:
227 //
228 // Coef for Y = (1 - kBt2100R - kBt2100B) / kBt2100G = 1
229 // Coef for U = kBt2100B * kBt2100Cb / kBt2100G = kBt2100GCb = ~0.1645
230 // Coef for V = kBt2100R * kBt2100Cr / kBt2100G = kBt2100GCr = ~0.5713
231
232 static const float kBt2100GCb = kBt2100B * kBt2100Cb / kBt2100G;
233 static const float kBt2100GCr = kBt2100R * kBt2100Cr / kBt2100G;
234
bt2100YuvToRgb(Color e_gamma)235 Color bt2100YuvToRgb(Color e_gamma) {
236 return {{{clampPixelFloat(e_gamma.y + kBt2100Cr * e_gamma.v),
237 clampPixelFloat(e_gamma.y - kBt2100GCb * e_gamma.u - kBt2100GCr * e_gamma.v),
238 clampPixelFloat(e_gamma.y + kBt2100Cb * e_gamma.u)}}};
239 }
240
241 // See ITU-R BT.2100-2, Table 5, HLG Reference OETF.
242 static const float kHlgA = 0.17883277f, kHlgB = 0.28466892f, kHlgC = 0.55991073f;
243
hlgOetf(float e)244 float hlgOetf(float e) {
245 if (e <= 1.0f / 12.0f) {
246 return sqrt(3.0f * e);
247 } else {
248 return kHlgA * log(12.0f * e - kHlgB) + kHlgC;
249 }
250 }
251
hlgOetf(Color e)252 Color hlgOetf(Color e) { return {{{hlgOetf(e.r), hlgOetf(e.g), hlgOetf(e.b)}}}; }
253
hlgOetfLUT(float e)254 float hlgOetfLUT(float e) {
255 int32_t value = static_cast<int32_t>(e * (kHlgOETFNumEntries - 1) + 0.5);
256 // TODO() : Remove once conversion modules have appropriate clamping in place
257 value = CLIP3(value, 0, kHlgOETFNumEntries - 1);
258 static LookUpTable kHlgLut(kHlgOETFNumEntries, static_cast<float (*)(float)>(hlgOetf));
259 return kHlgLut.getTable()[value];
260 }
261
hlgOetfLUT(Color e)262 Color hlgOetfLUT(Color e) { return {{{hlgOetfLUT(e.r), hlgOetfLUT(e.g), hlgOetfLUT(e.b)}}}; }
263
264 // See ITU-R BT.2100-2, Table 5, HLG Reference EOTF.
hlgInvOetf(float e_gamma)265 float hlgInvOetf(float e_gamma) {
266 if (e_gamma <= 0.5f) {
267 return pow(e_gamma, 2.0f) / 3.0f;
268 } else {
269 return (exp((e_gamma - kHlgC) / kHlgA) + kHlgB) / 12.0f;
270 }
271 }
272
hlgInvOetf(Color e_gamma)273 Color hlgInvOetf(Color e_gamma) {
274 return {{{hlgInvOetf(e_gamma.r), hlgInvOetf(e_gamma.g), hlgInvOetf(e_gamma.b)}}};
275 }
276
hlgInvOetfLUT(float e_gamma)277 float hlgInvOetfLUT(float e_gamma) {
278 int32_t value = static_cast<int32_t>(e_gamma * (kHlgInvOETFNumEntries - 1) + 0.5);
279 // TODO() : Remove once conversion modules have appropriate clamping in place
280 value = CLIP3(value, 0, kHlgInvOETFNumEntries - 1);
281 static LookUpTable kHlgInvLut(kHlgInvOETFNumEntries, static_cast<float (*)(float)>(hlgInvOetf));
282 return kHlgInvLut.getTable()[value];
283 }
284
hlgInvOetfLUT(Color e_gamma)285 Color hlgInvOetfLUT(Color e_gamma) {
286 return {{{hlgInvOetfLUT(e_gamma.r), hlgInvOetfLUT(e_gamma.g), hlgInvOetfLUT(e_gamma.b)}}};
287 }
288
289 // See ITU-R BT.2100-2, Table 5, Note 5f
290 // Gamma = 1.2 + 0.42 * log(kHlgMaxNits / 1000)
291 static const float kOotfGamma = 1.2f;
292
293 // See ITU-R BT.2100-2, Table 5, HLG Reference OOTF
hlgOotf(Color e,LuminanceFn luminance)294 Color hlgOotf(Color e, LuminanceFn luminance) {
295 float y = luminance(e);
296 return e * std::pow(y, kOotfGamma - 1.0f);
297 }
298
hlgOotfApprox(Color e,LuminanceFn luminance)299 Color hlgOotfApprox(Color e, [[maybe_unused]] LuminanceFn luminance) {
300 return {{{std::pow(e.r, kOotfGamma), std::pow(e.g, kOotfGamma), std::pow(e.b, kOotfGamma)}}};
301 }
302
303 // See ITU-R BT.2100-2, Table 5, Note 5i
hlgInverseOotf(Color e,LuminanceFn luminance)304 Color hlgInverseOotf(Color e, LuminanceFn luminance) {
305 float y = luminance(e);
306 return e * std::pow(y, (1.0f / kOotfGamma) - 1.0f);
307 }
308
hlgInverseOotfApprox(Color e)309 Color hlgInverseOotfApprox(Color e) {
310 return {{{std::pow(e.r, 1.0f / kOotfGamma), std::pow(e.g, 1.0f / kOotfGamma),
311 std::pow(e.b, 1.0f / kOotfGamma)}}};
312 }
313
314 // See ITU-R BT.2100-2, Table 4, Reference PQ OETF.
315 static const float kPqM1 = 2610.0f / 16384.0f, kPqM2 = 2523.0f / 4096.0f * 128.0f;
316 static const float kPqC1 = 3424.0f / 4096.0f, kPqC2 = 2413.0f / 4096.0f * 32.0f,
317 kPqC3 = 2392.0f / 4096.0f * 32.0f;
318
pqOetf(float e)319 float pqOetf(float e) {
320 if (e <= 0.0f) return 0.0f;
321 return pow((kPqC1 + kPqC2 * pow(e, kPqM1)) / (1 + kPqC3 * pow(e, kPqM1)), kPqM2);
322 }
323
pqOetf(Color e)324 Color pqOetf(Color e) { return {{{pqOetf(e.r), pqOetf(e.g), pqOetf(e.b)}}}; }
325
pqOetfLUT(float e)326 float pqOetfLUT(float e) {
327 int32_t value = static_cast<int32_t>(e * (kPqOETFNumEntries - 1) + 0.5);
328 // TODO() : Remove once conversion modules have appropriate clamping in place
329 value = CLIP3(value, 0, kPqOETFNumEntries - 1);
330 static LookUpTable kPqLut(kPqOETFNumEntries, static_cast<float (*)(float)>(pqOetf));
331 return kPqLut.getTable()[value];
332 }
333
pqOetfLUT(Color e)334 Color pqOetfLUT(Color e) { return {{{pqOetfLUT(e.r), pqOetfLUT(e.g), pqOetfLUT(e.b)}}}; }
335
pqInvOetf(float e_gamma)336 float pqInvOetf(float e_gamma) {
337 float val = pow(e_gamma, (1 / kPqM2));
338 return pow((((std::max)(val - kPqC1, 0.0f)) / (kPqC2 - kPqC3 * val)), 1 / kPqM1);
339 }
340
pqInvOetf(Color e_gamma)341 Color pqInvOetf(Color e_gamma) {
342 return {{{pqInvOetf(e_gamma.r), pqInvOetf(e_gamma.g), pqInvOetf(e_gamma.b)}}};
343 }
344
pqInvOetfLUT(float e_gamma)345 float pqInvOetfLUT(float e_gamma) {
346 int32_t value = static_cast<int32_t>(e_gamma * (kPqInvOETFNumEntries - 1) + 0.5);
347 // TODO() : Remove once conversion modules have appropriate clamping in place
348 value = CLIP3(value, 0, kPqInvOETFNumEntries - 1);
349 static LookUpTable kPqInvLut(kPqInvOETFNumEntries, static_cast<float (*)(float)>(pqInvOetf));
350 return kPqInvLut.getTable()[value];
351 }
352
pqInvOetfLUT(Color e_gamma)353 Color pqInvOetfLUT(Color e_gamma) {
354 return {{{pqInvOetfLUT(e_gamma.r), pqInvOetfLUT(e_gamma.g), pqInvOetfLUT(e_gamma.b)}}};
355 }
356
357 ////////////////////////////////////////////////////////////////////////////////
358 // Color access functions
359
getYuv4abPixel(uhdr_raw_image_t * image,size_t x,size_t y,int h_factor,int v_factor)360 Color getYuv4abPixel(uhdr_raw_image_t* image, size_t x, size_t y, int h_factor, int v_factor) {
361 uint8_t* luma_data = reinterpret_cast<uint8_t*>(image->planes[UHDR_PLANE_Y]);
362 size_t luma_stride = image->stride[UHDR_PLANE_Y];
363 uint8_t* cb_data = reinterpret_cast<uint8_t*>(image->planes[UHDR_PLANE_U]);
364 size_t cb_stride = image->stride[UHDR_PLANE_U];
365 uint8_t* cr_data = reinterpret_cast<uint8_t*>(image->planes[UHDR_PLANE_V]);
366 size_t cr_stride = image->stride[UHDR_PLANE_V];
367
368 size_t pixel_y_idx = x + y * luma_stride;
369 size_t pixel_cb_idx = x / h_factor + (y / v_factor) * cb_stride;
370 size_t pixel_cr_idx = x / h_factor + (y / v_factor) * cr_stride;
371
372 uint8_t y_uint = luma_data[pixel_y_idx];
373 uint8_t u_uint = cb_data[pixel_cb_idx];
374 uint8_t v_uint = cr_data[pixel_cr_idx];
375
376 // 128 bias for UV given we are using jpeglib; see:
377 // https://github.com/kornelski/libjpeg/blob/master/structure.doc
378 return {
379 {{static_cast<float>(y_uint) * (1 / 255.0f), static_cast<float>(u_uint - 128) * (1 / 255.0f),
380 static_cast<float>(v_uint - 128) * (1 / 255.0f)}}};
381 }
382
getYuv444Pixel(uhdr_raw_image_t * image,size_t x,size_t y)383 Color getYuv444Pixel(uhdr_raw_image_t* image, size_t x, size_t y) {
384 return getYuv4abPixel(image, x, y, 1, 1);
385 }
386
getYuv422Pixel(uhdr_raw_image_t * image,size_t x,size_t y)387 Color getYuv422Pixel(uhdr_raw_image_t* image, size_t x, size_t y) {
388 return getYuv4abPixel(image, x, y, 2, 1);
389 }
390
getYuv420Pixel(uhdr_raw_image_t * image,size_t x,size_t y)391 Color getYuv420Pixel(uhdr_raw_image_t* image, size_t x, size_t y) {
392 return getYuv4abPixel(image, x, y, 2, 2);
393 }
394
getYuv400Pixel(uhdr_raw_image_t * image,size_t x,size_t y)395 Color getYuv400Pixel(uhdr_raw_image_t* image, size_t x, size_t y) {
396 uint8_t* luma_data = reinterpret_cast<uint8_t*>(image->planes[UHDR_PLANE_Y]);
397 size_t luma_stride = image->stride[UHDR_PLANE_Y];
398 size_t pixel_y_idx = x + y * luma_stride;
399 uint8_t y_uint = luma_data[pixel_y_idx];
400
401 return {{{static_cast<float>(y_uint) * (1 / 255.0f), 0.f, 0.f}}};
402 }
403
getYuv444Pixel10bit(uhdr_raw_image_t * image,size_t x,size_t y)404 Color getYuv444Pixel10bit(uhdr_raw_image_t* image, size_t x, size_t y) {
405 uint16_t* luma_data = reinterpret_cast<uint16_t*>(image->planes[UHDR_PLANE_Y]);
406 size_t luma_stride = image->stride[UHDR_PLANE_Y];
407 uint16_t* cb_data = reinterpret_cast<uint16_t*>(image->planes[UHDR_PLANE_U]);
408 size_t cb_stride = image->stride[UHDR_PLANE_U];
409 uint16_t* cr_data = reinterpret_cast<uint16_t*>(image->planes[UHDR_PLANE_V]);
410 size_t cr_stride = image->stride[UHDR_PLANE_V];
411
412 size_t pixel_y_idx = y * luma_stride + x;
413 size_t pixel_u_idx = y * cb_stride + x;
414 size_t pixel_v_idx = y * cr_stride + x;
415
416 uint16_t y_uint = luma_data[pixel_y_idx];
417 uint16_t u_uint = cb_data[pixel_u_idx];
418 uint16_t v_uint = cr_data[pixel_v_idx];
419
420 if (image->range == UHDR_CR_FULL_RANGE) {
421 return {{{static_cast<float>(y_uint) / 1023.0f, static_cast<float>(u_uint) / 1023.0f - 0.5f,
422 static_cast<float>(v_uint) / 1023.0f - 0.5f}}};
423 }
424
425 // Conversions include taking narrow-range into account.
426 return {{{static_cast<float>(y_uint - 64) * (1 / 876.0f),
427 static_cast<float>(u_uint - 64) * (1 / 896.0f) - 0.5f,
428 static_cast<float>(v_uint - 64) * (1 / 896.0f) - 0.5f}}};
429 }
430
getP010Pixel(uhdr_raw_image_t * image,size_t x,size_t y)431 Color getP010Pixel(uhdr_raw_image_t* image, size_t x, size_t y) {
432 uint16_t* luma_data = reinterpret_cast<uint16_t*>(image->planes[UHDR_PLANE_Y]);
433 size_t luma_stride = image->stride[UHDR_PLANE_Y];
434 uint16_t* chroma_data = reinterpret_cast<uint16_t*>(image->planes[UHDR_PLANE_UV]);
435 size_t chroma_stride = image->stride[UHDR_PLANE_UV];
436
437 size_t pixel_y_idx = y * luma_stride + x;
438 size_t pixel_u_idx = (y >> 1) * chroma_stride + (x & ~0x1);
439 size_t pixel_v_idx = pixel_u_idx + 1;
440
441 uint16_t y_uint = luma_data[pixel_y_idx] >> 6;
442 uint16_t u_uint = chroma_data[pixel_u_idx] >> 6;
443 uint16_t v_uint = chroma_data[pixel_v_idx] >> 6;
444
445 if (image->range == UHDR_CR_FULL_RANGE) {
446 return {{{static_cast<float>(y_uint) / 1023.0f, static_cast<float>(u_uint) / 1023.0f - 0.5f,
447 static_cast<float>(v_uint) / 1023.0f - 0.5f}}};
448 }
449
450 // Conversions include taking narrow-range into account.
451 return {{{static_cast<float>(y_uint - 64) * (1 / 876.0f),
452 static_cast<float>(u_uint - 64) * (1 / 896.0f) - 0.5f,
453 static_cast<float>(v_uint - 64) * (1 / 896.0f) - 0.5f}}};
454 }
455
getRgb888Pixel(uhdr_raw_image_t * image,size_t x,size_t y)456 Color getRgb888Pixel(uhdr_raw_image_t* image, size_t x, size_t y) {
457 uint8_t* rgbData = static_cast<uint8_t*>(image->planes[UHDR_PLANE_PACKED]);
458 unsigned int srcStride = image->stride[UHDR_PLANE_PACKED];
459 size_t offset = x * 3 + y * srcStride * 3;
460 Color pixel;
461 pixel.r = float(rgbData[offset]);
462 pixel.g = float(rgbData[offset + 1]);
463 pixel.b = float(rgbData[offset + 2]);
464 return pixel / 255.0f;
465 }
466
getRgba8888Pixel(uhdr_raw_image_t * image,size_t x,size_t y)467 Color getRgba8888Pixel(uhdr_raw_image_t* image, size_t x, size_t y) {
468 uint32_t* rgbData = static_cast<uint32_t*>(image->planes[UHDR_PLANE_PACKED]);
469 unsigned int srcStride = image->stride[UHDR_PLANE_PACKED];
470
471 Color pixel;
472 pixel.r = float(rgbData[x + y * srcStride] & 0xff);
473 pixel.g = float((rgbData[x + y * srcStride] >> 8) & 0xff);
474 pixel.b = float((rgbData[x + y * srcStride] >> 16) & 0xff);
475 return pixel / 255.0f;
476 }
477
getRgba1010102Pixel(uhdr_raw_image_t * image,size_t x,size_t y)478 Color getRgba1010102Pixel(uhdr_raw_image_t* image, size_t x, size_t y) {
479 uint32_t* rgbData = static_cast<uint32_t*>(image->planes[UHDR_PLANE_PACKED]);
480 unsigned int srcStride = image->stride[UHDR_PLANE_PACKED];
481
482 Color pixel;
483 pixel.r = float(rgbData[x + y * srcStride] & 0x3ff);
484 pixel.g = float((rgbData[x + y * srcStride] >> 10) & 0x3ff);
485 pixel.b = float((rgbData[x + y * srcStride] >> 20) & 0x3ff);
486 return pixel / 1023.0f;
487 }
488
getRgbaF16Pixel(uhdr_raw_image_t * image,size_t x,size_t y)489 Color getRgbaF16Pixel(uhdr_raw_image_t* image, size_t x, size_t y) {
490 uint64_t* rgbData = static_cast<uint64_t*>(image->planes[UHDR_PLANE_PACKED]);
491 unsigned int srcStride = image->stride[UHDR_PLANE_PACKED];
492
493 Color pixel;
494 pixel.r = halfToFloat(rgbData[x + y * srcStride] & 0xffff);
495 pixel.g = halfToFloat((rgbData[x + y * srcStride] >> 16) & 0xffff);
496 pixel.b = halfToFloat((rgbData[x + y * srcStride] >> 32) & 0xffff);
497 return sanitizePixel(pixel);
498 }
499
samplePixels(uhdr_raw_image_t * image,size_t map_scale_factor,size_t x,size_t y,GetPixelFn get_pixel_fn)500 static Color samplePixels(uhdr_raw_image_t* image, size_t map_scale_factor, size_t x, size_t y,
501 GetPixelFn get_pixel_fn) {
502 Color e = {{{0.0f, 0.0f, 0.0f}}};
503 for (size_t dy = 0; dy < map_scale_factor; ++dy) {
504 for (size_t dx = 0; dx < map_scale_factor; ++dx) {
505 e += get_pixel_fn(image, x * map_scale_factor + dx, y * map_scale_factor + dy);
506 }
507 }
508
509 return e / static_cast<float>(map_scale_factor * map_scale_factor);
510 }
511
sampleYuv444(uhdr_raw_image_t * image,size_t map_scale_factor,size_t x,size_t y)512 Color sampleYuv444(uhdr_raw_image_t* image, size_t map_scale_factor, size_t x, size_t y) {
513 return samplePixels(image, map_scale_factor, x, y, getYuv444Pixel);
514 }
515
sampleYuv422(uhdr_raw_image_t * image,size_t map_scale_factor,size_t x,size_t y)516 Color sampleYuv422(uhdr_raw_image_t* image, size_t map_scale_factor, size_t x, size_t y) {
517 return samplePixels(image, map_scale_factor, x, y, getYuv422Pixel);
518 }
519
sampleYuv420(uhdr_raw_image_t * image,size_t map_scale_factor,size_t x,size_t y)520 Color sampleYuv420(uhdr_raw_image_t* image, size_t map_scale_factor, size_t x, size_t y) {
521 return samplePixels(image, map_scale_factor, x, y, getYuv420Pixel);
522 }
523
sampleP010(uhdr_raw_image_t * image,size_t map_scale_factor,size_t x,size_t y)524 Color sampleP010(uhdr_raw_image_t* image, size_t map_scale_factor, size_t x, size_t y) {
525 return samplePixels(image, map_scale_factor, x, y, getP010Pixel);
526 }
527
sampleYuv44410bit(uhdr_raw_image_t * image,size_t map_scale_factor,size_t x,size_t y)528 Color sampleYuv44410bit(uhdr_raw_image_t* image, size_t map_scale_factor, size_t x, size_t y) {
529 return samplePixels(image, map_scale_factor, x, y, getYuv444Pixel10bit);
530 }
531
sampleRgba8888(uhdr_raw_image_t * image,size_t map_scale_factor,size_t x,size_t y)532 Color sampleRgba8888(uhdr_raw_image_t* image, size_t map_scale_factor, size_t x, size_t y) {
533 return samplePixels(image, map_scale_factor, x, y, getRgba8888Pixel);
534 }
535
sampleRgba1010102(uhdr_raw_image_t * image,size_t map_scale_factor,size_t x,size_t y)536 Color sampleRgba1010102(uhdr_raw_image_t* image, size_t map_scale_factor, size_t x, size_t y) {
537 return samplePixels(image, map_scale_factor, x, y, getRgba1010102Pixel);
538 }
539
sampleRgbaF16(uhdr_raw_image_t * image,size_t map_scale_factor,size_t x,size_t y)540 Color sampleRgbaF16(uhdr_raw_image_t* image, size_t map_scale_factor, size_t x, size_t y) {
541 return samplePixels(image, map_scale_factor, x, y, getRgbaF16Pixel);
542 }
543
putRgba8888Pixel(uhdr_raw_image_t * image,size_t x,size_t y,Color & pixel)544 void putRgba8888Pixel(uhdr_raw_image_t* image, size_t x, size_t y, Color& pixel) {
545 uint32_t* rgbData = static_cast<uint32_t*>(image->planes[UHDR_PLANE_PACKED]);
546 unsigned int srcStride = image->stride[UHDR_PLANE_PACKED];
547
548 pixel *= 255.0f;
549 pixel += 0.5f;
550 pixel.r = CLIP3(pixel.r, 0.0f, 255.0f);
551 pixel.g = CLIP3(pixel.g, 0.0f, 255.0f);
552 pixel.b = CLIP3(pixel.b, 0.0f, 255.0f);
553
554 int32_t r0 = int32_t(pixel.r);
555 int32_t g0 = int32_t(pixel.g);
556 int32_t b0 = int32_t(pixel.b);
557 rgbData[x + y * srcStride] = r0 | (g0 << 8) | (b0 << 16) | (255 << 24); // Set alpha to 1.0
558 }
559
putRgb888Pixel(uhdr_raw_image_t * image,size_t x,size_t y,Color & pixel)560 void putRgb888Pixel(uhdr_raw_image_t* image, size_t x, size_t y, Color& pixel) {
561 uint8_t* rgbData = static_cast<uint8_t*>(image->planes[UHDR_PLANE_PACKED]);
562 unsigned int srcStride = image->stride[UHDR_PLANE_PACKED];
563 size_t offset = x * 3 + y * srcStride * 3;
564 pixel *= 255.0f;
565 pixel += 0.5f;
566 pixel.r = CLIP3(pixel.r, 0.0f, 255.0f);
567 pixel.g = CLIP3(pixel.g, 0.0f, 255.0f);
568 pixel.b = CLIP3(pixel.b, 0.0f, 255.0f);
569 rgbData[offset] = uint8_t(pixel.r);
570 rgbData[offset + 1] = uint8_t(pixel.r);
571 rgbData[offset + 2] = uint8_t(pixel.b);
572 }
573
putYuv400Pixel(uhdr_raw_image_t * image,size_t x,size_t y,Color & pixel)574 void putYuv400Pixel(uhdr_raw_image_t* image, size_t x, size_t y, Color& pixel) {
575 uint8_t* luma_data = reinterpret_cast<uint8_t*>(image->planes[UHDR_PLANE_Y]);
576 size_t luma_stride = image->stride[UHDR_PLANE_Y];
577
578 pixel *= 255.0f;
579 pixel += 0.5f;
580 pixel.y = CLIP3(pixel.y, 0.0f, 255.0f);
581
582 luma_data[x + y * luma_stride] = uint8_t(pixel.y);
583 }
584
putYuv444Pixel(uhdr_raw_image_t * image,size_t x,size_t y,Color & pixel)585 void putYuv444Pixel(uhdr_raw_image_t* image, size_t x, size_t y, Color& pixel) {
586 uint8_t* luma_data = reinterpret_cast<uint8_t*>(image->planes[UHDR_PLANE_Y]);
587 uint8_t* cb_data = reinterpret_cast<uint8_t*>(image->planes[UHDR_PLANE_U]);
588 uint8_t* cr_data = reinterpret_cast<uint8_t*>(image->planes[UHDR_PLANE_V]);
589 size_t luma_stride = image->stride[UHDR_PLANE_Y];
590 size_t cb_stride = image->stride[UHDR_PLANE_U];
591 size_t cr_stride = image->stride[UHDR_PLANE_V];
592
593 pixel *= 255.0f;
594 pixel += 0.5f;
595 pixel.y = CLIP3(pixel.y, 0.0f, 255.0f);
596 pixel.u = CLIP3(pixel.u, 0.0f, 255.0f);
597 pixel.v = CLIP3(pixel.v, 0.0f, 255.0f);
598
599 luma_data[x + y * luma_stride] = uint8_t(pixel.y);
600 cb_data[x + y * cb_stride] = uint8_t(pixel.u);
601 cr_data[x + y * cr_stride] = uint8_t(pixel.v);
602 }
603
604 ////////////////////////////////////////////////////////////////////////////////
605 // Color space conversions
606 // Sample, See,
607 // https://registry.khronos.org/DataFormat/specs/1.3/dataformat.1.3.html#_bt_709_bt_2020_primary_conversion_example
608
609 const std::array<float, 9> kBt709ToP3 = {0.822462f, 0.177537f, 0.000001f, 0.033194f, 0.966807f,
610 -0.000001f, 0.017083f, 0.072398f, 0.91052f};
611 const std::array<float, 9> kBt709ToBt2100 = {0.627404f, 0.329282f, 0.043314f, 0.069097f, 0.919541f,
612 0.011362f, 0.016392f, 0.088013f, 0.895595f};
613 const std::array<float, 9> kP3ToBt709 = {1.22494f, -0.22494f, 0.0f, -0.042057f, 1.042057f,
614 0.0f, -0.019638f, -0.078636f, 1.098274f};
615 const std::array<float, 9> kP3ToBt2100 = {0.753833f, 0.198597f, 0.04757f, 0.045744f, 0.941777f,
616 0.012479f, -0.00121f, 0.017601f, 0.983608f};
617 const std::array<float, 9> kBt2100ToBt709 = {1.660491f, -0.587641f, -0.07285f,
618 -0.124551f, 1.1329f, -0.008349f,
619 -0.018151f, -0.100579f, 1.11873f};
620 const std::array<float, 9> kBt2100ToP3 = {1.343578f, -0.282179f, -0.061399f, -0.065298f, 1.075788f,
621 -0.01049f, 0.002822f, -0.019598f, 1.016777f};
622
ConvertGamut(Color e,const std::array<float,9> & coeffs)623 Color ConvertGamut(Color e, const std::array<float, 9>& coeffs) {
624 return {{{coeffs[0] * e.r + coeffs[1] * e.g + coeffs[2] * e.b,
625 coeffs[3] * e.r + coeffs[4] * e.g + coeffs[5] * e.b,
626 coeffs[6] * e.r + coeffs[7] * e.g + coeffs[8] * e.b}}};
627 }
bt709ToP3(Color e)628 Color bt709ToP3(Color e) { return ConvertGamut(e, kBt709ToP3); }
bt709ToBt2100(Color e)629 Color bt709ToBt2100(Color e) { return ConvertGamut(e, kBt709ToBt2100); }
p3ToBt709(Color e)630 Color p3ToBt709(Color e) { return ConvertGamut(e, kP3ToBt709); }
p3ToBt2100(Color e)631 Color p3ToBt2100(Color e) { return ConvertGamut(e, kP3ToBt2100); }
bt2100ToBt709(Color e)632 Color bt2100ToBt709(Color e) { return ConvertGamut(e, kBt2100ToBt709); }
bt2100ToP3(Color e)633 Color bt2100ToP3(Color e) { return ConvertGamut(e, kBt2100ToP3); }
634
635 // All of these conversions are derived from the respective input YUV->RGB conversion followed by
636 // the RGB->YUV for the receiving encoding. They are consistent with the RGB<->YUV functions in
637 // gainmapmath.cpp, given that we use BT.709 encoding for sRGB and BT.601 encoding for Display-P3,
638 // to match DataSpace.
639
640 // Yuv Bt709 -> Yuv Bt601
641 // Y' = (1.0 * Y) + ( 0.101579 * U) + ( 0.196076 * V)
642 // U' = (0.0 * Y) + ( 0.989854 * U) + (-0.110653 * V)
643 // V' = (0.0 * Y) + (-0.072453 * U) + ( 0.983398 * V)
644 const std::array<float, 9> kYuvBt709ToBt601 = {
645 1.0f, 0.101579f, 0.196076f, 0.0f, 0.989854f, -0.110653f, 0.0f, -0.072453f, 0.983398f};
646
647 // Yuv Bt709 -> Yuv Bt2100
648 // Y' = (1.0 * Y) + (-0.016969 * U) + ( 0.096312 * V)
649 // U' = (0.0 * Y) + ( 0.995306 * U) + (-0.051192 * V)
650 // V' = (0.0 * Y) + ( 0.011507 * U) + ( 1.002637 * V)
651 const std::array<float, 9> kYuvBt709ToBt2100 = {
652 1.0f, -0.016969f, 0.096312f, 0.0f, 0.995306f, -0.051192f, 0.0f, 0.011507f, 1.002637f};
653
654 // Yuv Bt601 -> Yuv Bt709
655 // Y' = (1.0 * Y) + (-0.118188 * U) + (-0.212685 * V)
656 // U' = (0.0 * Y) + ( 1.018640 * U) + ( 0.114618 * V)
657 // V' = (0.0 * Y) + ( 0.075049 * U) + ( 1.025327 * V)
658 const std::array<float, 9> kYuvBt601ToBt709 = {
659 1.0f, -0.118188f, -0.212685f, 0.0f, 1.018640f, 0.114618f, 0.0f, 0.075049f, 1.025327f};
660
661 // Yuv Bt601 -> Yuv Bt2100
662 // Y' = (1.0 * Y) + (-0.128245 * U) + (-0.115879 * V)
663 // U' = (0.0 * Y) + ( 1.010016 * U) + ( 0.061592 * V)
664 // V' = (0.0 * Y) + ( 0.086969 * U) + ( 1.029350 * V)
665 const std::array<float, 9> kYuvBt601ToBt2100 = {
666 1.0f, -0.128245f, -0.115879, 0.0f, 1.010016f, 0.061592f, 0.0f, 0.086969f, 1.029350f};
667
668 // Yuv Bt2100 -> Yuv Bt709
669 // Y' = (1.0 * Y) + ( 0.018149 * U) + (-0.095132 * V)
670 // U' = (0.0 * Y) + ( 1.004123 * U) + ( 0.051267 * V)
671 // V' = (0.0 * Y) + (-0.011524 * U) + ( 0.996782 * V)
672 const std::array<float, 9> kYuvBt2100ToBt709 = {
673 1.0f, 0.018149f, -0.095132f, 0.0f, 1.004123f, 0.051267f, 0.0f, -0.011524f, 0.996782f};
674
675 // Yuv Bt2100 -> Yuv Bt601
676 // Y' = (1.0 * Y) + ( 0.117887 * U) + ( 0.105521 * V)
677 // U' = (0.0 * Y) + ( 0.995211 * U) + (-0.059549 * V)
678 // V' = (0.0 * Y) + (-0.084085 * U) + ( 0.976518 * V)
679 const std::array<float, 9> kYuvBt2100ToBt601 = {
680 1.0f, 0.117887f, 0.105521f, 0.0f, 0.995211f, -0.059549f, 0.0f, -0.084085f, 0.976518f};
681
yuvColorGamutConversion(Color e_gamma,const std::array<float,9> & coeffs)682 Color yuvColorGamutConversion(Color e_gamma, const std::array<float, 9>& coeffs) {
683 const float y = e_gamma.y * std::get<0>(coeffs) + e_gamma.u * std::get<1>(coeffs) +
684 e_gamma.v * std::get<2>(coeffs);
685 const float u = e_gamma.y * std::get<3>(coeffs) + e_gamma.u * std::get<4>(coeffs) +
686 e_gamma.v * std::get<5>(coeffs);
687 const float v = e_gamma.y * std::get<6>(coeffs) + e_gamma.u * std::get<7>(coeffs) +
688 e_gamma.v * std::get<8>(coeffs);
689 return {{{y, u, v}}};
690 }
691
transformYuv420(uhdr_raw_image_t * image,const std::array<float,9> & coeffs)692 void transformYuv420(uhdr_raw_image_t* image, const std::array<float, 9>& coeffs) {
693 for (size_t y = 0; y < image->h / 2; ++y) {
694 for (size_t x = 0; x < image->w / 2; ++x) {
695 Color yuv1 = getYuv420Pixel(image, x * 2, y * 2);
696 Color yuv2 = getYuv420Pixel(image, x * 2 + 1, y * 2);
697 Color yuv3 = getYuv420Pixel(image, x * 2, y * 2 + 1);
698 Color yuv4 = getYuv420Pixel(image, x * 2 + 1, y * 2 + 1);
699
700 yuv1 = yuvColorGamutConversion(yuv1, coeffs);
701 yuv2 = yuvColorGamutConversion(yuv2, coeffs);
702 yuv3 = yuvColorGamutConversion(yuv3, coeffs);
703 yuv4 = yuvColorGamutConversion(yuv4, coeffs);
704
705 Color new_uv = (yuv1 + yuv2 + yuv3 + yuv4) / 4.0f;
706
707 size_t pixel_y1_idx = x * 2 + y * 2 * image->stride[UHDR_PLANE_Y];
708 size_t pixel_y2_idx = (x * 2 + 1) + y * 2 * image->stride[UHDR_PLANE_Y];
709 size_t pixel_y3_idx = x * 2 + (y * 2 + 1) * image->stride[UHDR_PLANE_Y];
710 size_t pixel_y4_idx = (x * 2 + 1) + (y * 2 + 1) * image->stride[UHDR_PLANE_Y];
711
712 uint8_t& y1_uint = reinterpret_cast<uint8_t*>(image->planes[UHDR_PLANE_Y])[pixel_y1_idx];
713 uint8_t& y2_uint = reinterpret_cast<uint8_t*>(image->planes[UHDR_PLANE_Y])[pixel_y2_idx];
714 uint8_t& y3_uint = reinterpret_cast<uint8_t*>(image->planes[UHDR_PLANE_Y])[pixel_y3_idx];
715 uint8_t& y4_uint = reinterpret_cast<uint8_t*>(image->planes[UHDR_PLANE_Y])[pixel_y4_idx];
716
717 size_t pixel_u_idx = x + y * image->stride[UHDR_PLANE_U];
718 uint8_t& u_uint = reinterpret_cast<uint8_t*>(image->planes[UHDR_PLANE_U])[pixel_u_idx];
719
720 size_t pixel_v_idx = x + y * image->stride[UHDR_PLANE_V];
721 uint8_t& v_uint = reinterpret_cast<uint8_t*>(image->planes[UHDR_PLANE_V])[pixel_v_idx];
722
723 y1_uint = static_cast<uint8_t>(CLIP3((yuv1.y * 255.0f + 0.5f), 0, 255));
724 y2_uint = static_cast<uint8_t>(CLIP3((yuv2.y * 255.0f + 0.5f), 0, 255));
725 y3_uint = static_cast<uint8_t>(CLIP3((yuv3.y * 255.0f + 0.5f), 0, 255));
726 y4_uint = static_cast<uint8_t>(CLIP3((yuv4.y * 255.0f + 0.5f), 0, 255));
727
728 u_uint = static_cast<uint8_t>(CLIP3((new_uv.u * 255.0f + 128.0f + 0.5f), 0, 255));
729 v_uint = static_cast<uint8_t>(CLIP3((new_uv.v * 255.0f + 128.0f + 0.5f), 0, 255));
730 }
731 }
732 }
733
transformYuv444(uhdr_raw_image_t * image,const std::array<float,9> & coeffs)734 void transformYuv444(uhdr_raw_image_t* image, const std::array<float, 9>& coeffs) {
735 for (size_t y = 0; y < image->h; ++y) {
736 for (size_t x = 0; x < image->w; ++x) {
737 Color yuv = getYuv444Pixel(image, x, y);
738 yuv = yuvColorGamutConversion(yuv, coeffs);
739
740 size_t pixel_y_idx = x + y * image->stride[UHDR_PLANE_Y];
741 uint8_t& y1_uint = reinterpret_cast<uint8_t*>(image->planes[UHDR_PLANE_Y])[pixel_y_idx];
742
743 size_t pixel_u_idx = x + y * image->stride[UHDR_PLANE_U];
744 uint8_t& u_uint = reinterpret_cast<uint8_t*>(image->planes[UHDR_PLANE_U])[pixel_u_idx];
745
746 size_t pixel_v_idx = x + y * image->stride[UHDR_PLANE_V];
747 uint8_t& v_uint = reinterpret_cast<uint8_t*>(image->planes[UHDR_PLANE_V])[pixel_v_idx];
748
749 y1_uint = static_cast<uint8_t>(CLIP3((yuv.y * 255.0f + 0.5f), 0, 255));
750 u_uint = static_cast<uint8_t>(CLIP3((yuv.u * 255.0f + 128.0f + 0.5f), 0, 255));
751 v_uint = static_cast<uint8_t>(CLIP3((yuv.v * 255.0f + 128.0f + 0.5f), 0, 255));
752 }
753 }
754 }
755
756 ////////////////////////////////////////////////////////////////////////////////
757 // Gain map calculations
758
encodeGain(float y_sdr,float y_hdr,uhdr_gainmap_metadata_ext_t * metadata,int index)759 uint8_t encodeGain(float y_sdr, float y_hdr, uhdr_gainmap_metadata_ext_t* metadata, int index) {
760 return encodeGain(y_sdr, y_hdr, metadata, log2(metadata->min_content_boost[index]),
761 log2(metadata->max_content_boost[index]), index);
762 }
763
encodeGain(float y_sdr,float y_hdr,uhdr_gainmap_metadata_ext_t * metadata,float log2MinContentBoost,float log2MaxContentBoost,int index)764 uint8_t encodeGain(float y_sdr, float y_hdr, uhdr_gainmap_metadata_ext_t* metadata,
765 float log2MinContentBoost, float log2MaxContentBoost, int index) {
766 float gain = 1.0f;
767 if (y_sdr > 0.0f) {
768 gain = y_hdr / y_sdr;
769 }
770
771 if (gain < metadata->min_content_boost[index]) gain = metadata->min_content_boost[index];
772 if (gain > metadata->max_content_boost[index]) gain = metadata->max_content_boost[index];
773 float gain_normalized =
774 (log2(gain) - log2MinContentBoost) / (log2MaxContentBoost - log2MinContentBoost);
775 float gain_normalized_gamma = powf(gain_normalized, metadata->gamma[index]);
776 return static_cast<uint8_t>(gain_normalized_gamma * 255.0f);
777 }
778
computeGain(float sdr,float hdr)779 float computeGain(float sdr, float hdr) {
780 float gain = log2((hdr + kHdrOffset) / (sdr + kSdrOffset));
781 if (sdr < 2.f / 255.0f) {
782 // If sdr is zero and hdr is non zero, it can result in very large gain values. In compression -
783 // decompression process, if the same sdr pixel increases to 1, the hdr recovered pixel will
784 // blow out. Dont allow dark pixels to signal large gains.
785 gain = (std::min)(gain, 2.3f);
786 }
787 return gain;
788 }
789
affineMapGain(float gainlog2,float mingainlog2,float maxgainlog2,float gamma)790 uint8_t affineMapGain(float gainlog2, float mingainlog2, float maxgainlog2, float gamma) {
791 float mappedVal = (gainlog2 - mingainlog2) / (maxgainlog2 - mingainlog2);
792 if (gamma != 1.0f) mappedVal = pow(mappedVal, gamma);
793 mappedVal *= 255;
794 return CLIP3(mappedVal + 0.5f, 0, 255);
795 }
796
applyGain(Color e,float gain,uhdr_gainmap_metadata_ext_t * metadata)797 Color applyGain(Color e, float gain, uhdr_gainmap_metadata_ext_t* metadata) {
798 if (metadata->gamma[0] != 1.0f) gain = pow(gain, 1.0f / metadata->gamma[0]);
799 float logBoost = log2(metadata->min_content_boost[0]) * (1.0f - gain) +
800 log2(metadata->max_content_boost[0]) * gain;
801 float gainFactor = exp2(logBoost);
802 return ((e + metadata->offset_sdr[0]) * gainFactor) - metadata->offset_hdr[0];
803 }
804
applyGain(Color e,float gain,uhdr_gainmap_metadata_ext_t * metadata,float gainmapWeight)805 Color applyGain(Color e, float gain, uhdr_gainmap_metadata_ext_t* metadata, float gainmapWeight) {
806 if (metadata->gamma[0] != 1.0f) gain = pow(gain, 1.0f / metadata->gamma[0]);
807 float logBoost = log2(metadata->min_content_boost[0]) * (1.0f - gain) +
808 log2(metadata->max_content_boost[0]) * gain;
809 float gainFactor = exp2(logBoost * gainmapWeight);
810 return ((e + metadata->offset_sdr[0]) * gainFactor) - metadata->offset_hdr[0];
811 }
812
applyGainLUT(Color e,float gain,GainLUT & gainLUT,uhdr_gainmap_metadata_ext_t * metadata)813 Color applyGainLUT(Color e, float gain, GainLUT& gainLUT, uhdr_gainmap_metadata_ext_t* metadata) {
814 float gainFactor = gainLUT.getGainFactor(gain, 0);
815 return ((e + metadata->offset_sdr[0]) * gainFactor) - metadata->offset_hdr[0];
816 }
817
applyGain(Color e,Color gain,uhdr_gainmap_metadata_ext_t * metadata)818 Color applyGain(Color e, Color gain, uhdr_gainmap_metadata_ext_t* metadata) {
819 if (metadata->gamma[0] != 1.0f) gain.r = pow(gain.r, 1.0f / metadata->gamma[0]);
820 if (metadata->gamma[1] != 1.0f) gain.g = pow(gain.g, 1.0f / metadata->gamma[1]);
821 if (metadata->gamma[2] != 1.0f) gain.b = pow(gain.b, 1.0f / metadata->gamma[2]);
822 float logBoostR = log2(metadata->min_content_boost[0]) * (1.0f - gain.r) +
823 log2(metadata->max_content_boost[0]) * gain.r;
824 float logBoostG = log2(metadata->min_content_boost[1]) * (1.0f - gain.g) +
825 log2(metadata->max_content_boost[1]) * gain.g;
826 float logBoostB = log2(metadata->min_content_boost[2]) * (1.0f - gain.b) +
827 log2(metadata->max_content_boost[2]) * gain.b;
828 float gainFactorR = exp2(logBoostR);
829 float gainFactorG = exp2(logBoostG);
830 float gainFactorB = exp2(logBoostB);
831 return {{{((e.r + metadata->offset_sdr[0]) * gainFactorR) - metadata->offset_hdr[0],
832 ((e.g + metadata->offset_sdr[1]) * gainFactorG) - metadata->offset_hdr[1],
833 ((e.b + metadata->offset_sdr[2]) * gainFactorB) - metadata->offset_hdr[2]}}};
834 }
835
applyGain(Color e,Color gain,uhdr_gainmap_metadata_ext_t * metadata,float gainmapWeight)836 Color applyGain(Color e, Color gain, uhdr_gainmap_metadata_ext_t* metadata, float gainmapWeight) {
837 if (metadata->gamma[0] != 1.0f) gain.r = pow(gain.r, 1.0f / metadata->gamma[0]);
838 if (metadata->gamma[1] != 1.0f) gain.g = pow(gain.g, 1.0f / metadata->gamma[1]);
839 if (metadata->gamma[2] != 1.0f) gain.b = pow(gain.b, 1.0f / metadata->gamma[2]);
840 float logBoostR = log2(metadata->min_content_boost[0]) * (1.0f - gain.r) +
841 log2(metadata->max_content_boost[0]) * gain.r;
842 float logBoostG = log2(metadata->min_content_boost[1]) * (1.0f - gain.g) +
843 log2(metadata->max_content_boost[1]) * gain.g;
844 float logBoostB = log2(metadata->min_content_boost[2]) * (1.0f - gain.b) +
845 log2(metadata->max_content_boost[2]) * gain.b;
846 float gainFactorR = exp2(logBoostR * gainmapWeight);
847 float gainFactorG = exp2(logBoostG * gainmapWeight);
848 float gainFactorB = exp2(logBoostB * gainmapWeight);
849 return {{{((e.r + metadata->offset_sdr[0]) * gainFactorR) - metadata->offset_hdr[0],
850 ((e.g + metadata->offset_sdr[1]) * gainFactorG) - metadata->offset_hdr[1],
851 ((e.b + metadata->offset_sdr[2]) * gainFactorB) - metadata->offset_hdr[2]}}};
852 }
853
applyGainLUT(Color e,Color gain,GainLUT & gainLUT,uhdr_gainmap_metadata_ext_t * metadata)854 Color applyGainLUT(Color e, Color gain, GainLUT& gainLUT, uhdr_gainmap_metadata_ext_t* metadata) {
855 float gainFactorR = gainLUT.getGainFactor(gain.r, 0);
856 float gainFactorG = gainLUT.getGainFactor(gain.g, 1);
857 float gainFactorB = gainLUT.getGainFactor(gain.b, 2);
858 return {{{((e.r + metadata->offset_sdr[0]) * gainFactorR) - metadata->offset_hdr[0],
859 ((e.g + metadata->offset_sdr[1]) * gainFactorG) - metadata->offset_hdr[1],
860 ((e.b + metadata->offset_sdr[2]) * gainFactorB) - metadata->offset_hdr[2]}}};
861 }
862
863 // TODO: do we need something more clever for filtering either the map or images
864 // to generate the map?
865
clamp(const size_t & val,const size_t & low,const size_t & high)866 static size_t clamp(const size_t& val, const size_t& low, const size_t& high) {
867 return val < low ? low : (high < val ? high : val);
868 }
869
mapUintToFloat(uint8_t map_uint)870 static float mapUintToFloat(uint8_t map_uint) { return static_cast<float>(map_uint) / 255.0f; }
871
pythDistance(float x_diff,float y_diff)872 static float pythDistance(float x_diff, float y_diff) {
873 return sqrt(pow(x_diff, 2.0f) + pow(y_diff, 2.0f));
874 }
875
876 // TODO: If map_scale_factor is guaranteed to be an integer, then remove the following.
sampleMap(uhdr_raw_image_t * map,float map_scale_factor,size_t x,size_t y)877 float sampleMap(uhdr_raw_image_t* map, float map_scale_factor, size_t x, size_t y) {
878 float x_map = static_cast<float>(x) / map_scale_factor;
879 float y_map = static_cast<float>(y) / map_scale_factor;
880
881 size_t x_lower = static_cast<size_t>(floor(x_map));
882 size_t x_upper = x_lower + 1;
883 size_t y_lower = static_cast<size_t>(floor(y_map));
884 size_t y_upper = y_lower + 1;
885
886 x_lower = clamp(x_lower, 0, map->w - 1);
887 x_upper = clamp(x_upper, 0, map->w - 1);
888 y_lower = clamp(y_lower, 0, map->h - 1);
889 y_upper = clamp(y_upper, 0, map->h - 1);
890
891 // Use Shepard's method for inverse distance weighting. For more information:
892 // en.wikipedia.org/wiki/Inverse_distance_weighting#Shepard's_method
893 uint8_t* data = reinterpret_cast<uint8_t*>(map->planes[UHDR_PLANE_Y]);
894 size_t stride = map->stride[UHDR_PLANE_Y];
895
896 float e1 = mapUintToFloat(data[x_lower + y_lower * stride]);
897 float e1_dist =
898 pythDistance(x_map - static_cast<float>(x_lower), y_map - static_cast<float>(y_lower));
899 if (e1_dist == 0.0f) return e1;
900
901 float e2 = mapUintToFloat(data[x_lower + y_upper * stride]);
902 float e2_dist =
903 pythDistance(x_map - static_cast<float>(x_lower), y_map - static_cast<float>(y_upper));
904 if (e2_dist == 0.0f) return e2;
905
906 float e3 = mapUintToFloat(data[x_upper + y_lower * stride]);
907 float e3_dist =
908 pythDistance(x_map - static_cast<float>(x_upper), y_map - static_cast<float>(y_lower));
909 if (e3_dist == 0.0f) return e3;
910
911 float e4 = mapUintToFloat(data[x_upper + y_upper * stride]);
912 float e4_dist =
913 pythDistance(x_map - static_cast<float>(x_upper), y_map - static_cast<float>(y_upper));
914 if (e4_dist == 0.0f) return e2;
915
916 float e1_weight = 1.0f / e1_dist;
917 float e2_weight = 1.0f / e2_dist;
918 float e3_weight = 1.0f / e3_dist;
919 float e4_weight = 1.0f / e4_dist;
920 float total_weight = e1_weight + e2_weight + e3_weight + e4_weight;
921
922 return e1 * (e1_weight / total_weight) + e2 * (e2_weight / total_weight) +
923 e3 * (e3_weight / total_weight) + e4 * (e4_weight / total_weight);
924 }
925
sampleMap(uhdr_raw_image_t * map,size_t map_scale_factor,size_t x,size_t y,ShepardsIDW & weightTables)926 float sampleMap(uhdr_raw_image_t* map, size_t map_scale_factor, size_t x, size_t y,
927 ShepardsIDW& weightTables) {
928 // TODO: If map_scale_factor is guaranteed to be an integer power of 2, then optimize the
929 // following by computing log2(map_scale_factor) once and then using >> log2(map_scale_factor)
930 size_t x_lower = x / map_scale_factor;
931 size_t x_upper = x_lower + 1;
932 size_t y_lower = y / map_scale_factor;
933 size_t y_upper = y_lower + 1;
934
935 x_lower = std::min(x_lower, (size_t)map->w - 1);
936 x_upper = std::min(x_upper, (size_t)map->w - 1);
937 y_lower = std::min(y_lower, (size_t)map->h - 1);
938 y_upper = std::min(y_upper, (size_t)map->h - 1);
939
940 uint8_t* data = reinterpret_cast<uint8_t*>(map->planes[UHDR_PLANE_Y]);
941 size_t stride = map->stride[UHDR_PLANE_Y];
942 float e1 = mapUintToFloat(data[x_lower + y_lower * stride]);
943 float e2 = mapUintToFloat(data[x_lower + y_upper * stride]);
944 float e3 = mapUintToFloat(data[x_upper + y_lower * stride]);
945 float e4 = mapUintToFloat(data[x_upper + y_upper * stride]);
946
947 // TODO: If map_scale_factor is guaranteed to be an integer power of 2, then optimize the
948 // following by using & (map_scale_factor - 1)
949 size_t offset_x = x % map_scale_factor;
950 size_t offset_y = y % map_scale_factor;
951
952 float* weights = weightTables.mWeights;
953 if (x_lower == x_upper && y_lower == y_upper)
954 weights = weightTables.mWeightsC;
955 else if (x_lower == x_upper)
956 weights = weightTables.mWeightsNR;
957 else if (y_lower == y_upper)
958 weights = weightTables.mWeightsNB;
959 weights += offset_y * map_scale_factor * 4 + offset_x * 4;
960
961 return e1 * weights[0] + e2 * weights[1] + e3 * weights[2] + e4 * weights[3];
962 }
963
sampleMap3Channel(uhdr_raw_image_t * map,float map_scale_factor,size_t x,size_t y,bool has_alpha)964 Color sampleMap3Channel(uhdr_raw_image_t* map, float map_scale_factor, size_t x, size_t y,
965 bool has_alpha) {
966 float x_map = static_cast<float>(x) / map_scale_factor;
967 float y_map = static_cast<float>(y) / map_scale_factor;
968
969 size_t x_lower = static_cast<size_t>(floor(x_map));
970 size_t x_upper = x_lower + 1;
971 size_t y_lower = static_cast<size_t>(floor(y_map));
972 size_t y_upper = y_lower + 1;
973
974 x_lower = std::min(x_lower, (size_t)map->w - 1);
975 x_upper = std::min(x_upper, (size_t)map->w - 1);
976 y_lower = std::min(y_lower, (size_t)map->h - 1);
977 y_upper = std::min(y_upper, (size_t)map->h - 1);
978
979 int factor = has_alpha ? 4 : 3;
980
981 uint8_t* data = reinterpret_cast<uint8_t*>(map->planes[UHDR_PLANE_PACKED]);
982 size_t stride = map->stride[UHDR_PLANE_PACKED];
983
984 float r1 = mapUintToFloat(data[(x_lower + y_lower * stride) * factor]);
985 float r2 = mapUintToFloat(data[(x_lower + y_upper * stride) * factor]);
986 float r3 = mapUintToFloat(data[(x_upper + y_lower * stride) * factor]);
987 float r4 = mapUintToFloat(data[(x_upper + y_upper * stride) * factor]);
988
989 float g1 = mapUintToFloat(data[(x_lower + y_lower * stride) * factor + 1]);
990 float g2 = mapUintToFloat(data[(x_lower + y_upper * stride) * factor + 1]);
991 float g3 = mapUintToFloat(data[(x_upper + y_lower * stride) * factor + 1]);
992 float g4 = mapUintToFloat(data[(x_upper + y_upper * stride) * factor + 1]);
993
994 float b1 = mapUintToFloat(data[(x_lower + y_lower * stride) * factor + 2]);
995 float b2 = mapUintToFloat(data[(x_lower + y_upper * stride) * factor + 2]);
996 float b3 = mapUintToFloat(data[(x_upper + y_lower * stride) * factor + 2]);
997 float b4 = mapUintToFloat(data[(x_upper + y_upper * stride) * factor + 2]);
998
999 Color rgb1 = {{{r1, g1, b1}}};
1000 Color rgb2 = {{{r2, g2, b2}}};
1001 Color rgb3 = {{{r3, g3, b3}}};
1002 Color rgb4 = {{{r4, g4, b4}}};
1003
1004 // Use Shepard's method for inverse distance weighting. For more information:
1005 // en.wikipedia.org/wiki/Inverse_distance_weighting#Shepard's_method
1006 float e1_dist =
1007 pythDistance(x_map - static_cast<float>(x_lower), y_map - static_cast<float>(y_lower));
1008 if (e1_dist == 0.0f) return rgb1;
1009
1010 float e2_dist =
1011 pythDistance(x_map - static_cast<float>(x_lower), y_map - static_cast<float>(y_upper));
1012 if (e2_dist == 0.0f) return rgb2;
1013
1014 float e3_dist =
1015 pythDistance(x_map - static_cast<float>(x_upper), y_map - static_cast<float>(y_lower));
1016 if (e3_dist == 0.0f) return rgb3;
1017
1018 float e4_dist =
1019 pythDistance(x_map - static_cast<float>(x_upper), y_map - static_cast<float>(y_upper));
1020 if (e4_dist == 0.0f) return rgb4;
1021
1022 float e1_weight = 1.0f / e1_dist;
1023 float e2_weight = 1.0f / e2_dist;
1024 float e3_weight = 1.0f / e3_dist;
1025 float e4_weight = 1.0f / e4_dist;
1026 float total_weight = e1_weight + e2_weight + e3_weight + e4_weight;
1027
1028 return rgb1 * (e1_weight / total_weight) + rgb2 * (e2_weight / total_weight) +
1029 rgb3 * (e3_weight / total_weight) + rgb4 * (e4_weight / total_weight);
1030 }
1031
sampleMap3Channel(uhdr_raw_image_t * map,size_t map_scale_factor,size_t x,size_t y,ShepardsIDW & weightTables,bool has_alpha)1032 Color sampleMap3Channel(uhdr_raw_image_t* map, size_t map_scale_factor, size_t x, size_t y,
1033 ShepardsIDW& weightTables, bool has_alpha) {
1034 // TODO: If map_scale_factor is guaranteed to be an integer power of 2, then optimize the
1035 // following by computing log2(map_scale_factor) once and then using >> log2(map_scale_factor)
1036 size_t x_lower = x / map_scale_factor;
1037 size_t x_upper = x_lower + 1;
1038 size_t y_lower = y / map_scale_factor;
1039 size_t y_upper = y_lower + 1;
1040
1041 x_lower = std::min(x_lower, (size_t)map->w - 1);
1042 x_upper = std::min(x_upper, (size_t)map->w - 1);
1043 y_lower = std::min(y_lower, (size_t)map->h - 1);
1044 y_upper = std::min(y_upper, (size_t)map->h - 1);
1045
1046 int factor = has_alpha ? 4 : 3;
1047
1048 uint8_t* data = reinterpret_cast<uint8_t*>(map->planes[UHDR_PLANE_PACKED]);
1049 size_t stride = map->stride[UHDR_PLANE_PACKED];
1050
1051 float r1 = mapUintToFloat(data[(x_lower + y_lower * stride) * factor]);
1052 float r2 = mapUintToFloat(data[(x_lower + y_upper * stride) * factor]);
1053 float r3 = mapUintToFloat(data[(x_upper + y_lower * stride) * factor]);
1054 float r4 = mapUintToFloat(data[(x_upper + y_upper * stride) * factor]);
1055
1056 float g1 = mapUintToFloat(data[(x_lower + y_lower * stride) * factor + 1]);
1057 float g2 = mapUintToFloat(data[(x_lower + y_upper * stride) * factor + 1]);
1058 float g3 = mapUintToFloat(data[(x_upper + y_lower * stride) * factor + 1]);
1059 float g4 = mapUintToFloat(data[(x_upper + y_upper * stride) * factor + 1]);
1060
1061 float b1 = mapUintToFloat(data[(x_lower + y_lower * stride) * factor + 2]);
1062 float b2 = mapUintToFloat(data[(x_lower + y_upper * stride) * factor + 2]);
1063 float b3 = mapUintToFloat(data[(x_upper + y_lower * stride) * factor + 2]);
1064 float b4 = mapUintToFloat(data[(x_upper + y_upper * stride) * factor + 2]);
1065
1066 Color rgb1 = {{{r1, g1, b1}}};
1067 Color rgb2 = {{{r2, g2, b2}}};
1068 Color rgb3 = {{{r3, g3, b3}}};
1069 Color rgb4 = {{{r4, g4, b4}}};
1070
1071 // TODO: If map_scale_factor is guaranteed to be an integer power of 2, then optimize the
1072 // following by using & (map_scale_factor - 1)
1073 size_t offset_x = x % map_scale_factor;
1074 size_t offset_y = y % map_scale_factor;
1075
1076 float* weights = weightTables.mWeights;
1077 if (x_lower == x_upper && y_lower == y_upper)
1078 weights = weightTables.mWeightsC;
1079 else if (x_lower == x_upper)
1080 weights = weightTables.mWeightsNR;
1081 else if (y_lower == y_upper)
1082 weights = weightTables.mWeightsNB;
1083 weights += offset_y * map_scale_factor * 4 + offset_x * 4;
1084
1085 return rgb1 * weights[0] + rgb2 * weights[1] + rgb3 * weights[2] + rgb4 * weights[3];
1086 }
1087
1088 ////////////////////////////////////////////////////////////////////////////////
1089 // function selectors
1090
1091 // TODO: confirm we always want to convert like this before calculating
1092 // luminance.
getGamutConversionFn(uhdr_color_gamut_t dst_gamut,uhdr_color_gamut_t src_gamut)1093 ColorTransformFn getGamutConversionFn(uhdr_color_gamut_t dst_gamut, uhdr_color_gamut_t src_gamut) {
1094 switch (dst_gamut) {
1095 case UHDR_CG_BT_709:
1096 switch (src_gamut) {
1097 case UHDR_CG_BT_709:
1098 return identityConversion;
1099 case UHDR_CG_DISPLAY_P3:
1100 return p3ToBt709;
1101 case UHDR_CG_BT_2100:
1102 return bt2100ToBt709;
1103 case UHDR_CG_UNSPECIFIED:
1104 return nullptr;
1105 }
1106 break;
1107 case UHDR_CG_DISPLAY_P3:
1108 switch (src_gamut) {
1109 case UHDR_CG_BT_709:
1110 return bt709ToP3;
1111 case UHDR_CG_DISPLAY_P3:
1112 return identityConversion;
1113 case UHDR_CG_BT_2100:
1114 return bt2100ToP3;
1115 case UHDR_CG_UNSPECIFIED:
1116 return nullptr;
1117 }
1118 break;
1119 case UHDR_CG_BT_2100:
1120 switch (src_gamut) {
1121 case UHDR_CG_BT_709:
1122 return bt709ToBt2100;
1123 case UHDR_CG_DISPLAY_P3:
1124 return p3ToBt2100;
1125 case UHDR_CG_BT_2100:
1126 return identityConversion;
1127 case UHDR_CG_UNSPECIFIED:
1128 return nullptr;
1129 }
1130 break;
1131 case UHDR_CG_UNSPECIFIED:
1132 return nullptr;
1133 }
1134 return nullptr;
1135 }
1136
getYuvToRgbFn(uhdr_color_gamut_t gamut)1137 ColorTransformFn getYuvToRgbFn(uhdr_color_gamut_t gamut) {
1138 switch (gamut) {
1139 case UHDR_CG_BT_709:
1140 return srgbYuvToRgb;
1141 case UHDR_CG_DISPLAY_P3:
1142 return p3YuvToRgb;
1143 case UHDR_CG_BT_2100:
1144 return bt2100YuvToRgb;
1145 case UHDR_CG_UNSPECIFIED:
1146 return nullptr;
1147 }
1148 return nullptr;
1149 }
1150
getLuminanceFn(uhdr_color_gamut_t gamut)1151 LuminanceFn getLuminanceFn(uhdr_color_gamut_t gamut) {
1152 switch (gamut) {
1153 case UHDR_CG_BT_709:
1154 return srgbLuminance;
1155 case UHDR_CG_DISPLAY_P3:
1156 return p3Luminance;
1157 case UHDR_CG_BT_2100:
1158 return bt2100Luminance;
1159 case UHDR_CG_UNSPECIFIED:
1160 return nullptr;
1161 }
1162 return nullptr;
1163 }
1164
getInverseOetfFn(uhdr_color_transfer_t transfer)1165 ColorTransformFn getInverseOetfFn(uhdr_color_transfer_t transfer) {
1166 switch (transfer) {
1167 case UHDR_CT_LINEAR:
1168 return identityConversion;
1169 case UHDR_CT_HLG:
1170 #if USE_HLG_INVOETF_LUT
1171 return hlgInvOetfLUT;
1172 #else
1173 return hlgInvOetf;
1174 #endif
1175 case UHDR_CT_PQ:
1176 #if USE_PQ_INVOETF_LUT
1177 return pqInvOetfLUT;
1178 #else
1179 return pqInvOetf;
1180 #endif
1181 case UHDR_CT_SRGB:
1182 #if USE_SRGB_INVOETF_LUT
1183 return srgbInvOetfLUT;
1184 #else
1185 return srgbInvOetf;
1186 #endif
1187 case UHDR_CT_UNSPECIFIED:
1188 return nullptr;
1189 }
1190 return nullptr;
1191 }
1192
getOotfFn(uhdr_color_transfer_t transfer)1193 SceneToDisplayLuminanceFn getOotfFn(uhdr_color_transfer_t transfer) {
1194 switch (transfer) {
1195 case UHDR_CT_LINEAR:
1196 return identityOotf;
1197 case UHDR_CT_HLG:
1198 return hlgOotfApprox;
1199 case UHDR_CT_PQ:
1200 return identityOotf;
1201 case UHDR_CT_SRGB:
1202 return identityOotf;
1203 case UHDR_CT_UNSPECIFIED:
1204 return nullptr;
1205 }
1206 return nullptr;
1207 }
1208
getPixelFn(uhdr_img_fmt_t format)1209 GetPixelFn getPixelFn(uhdr_img_fmt_t format) {
1210 switch (format) {
1211 case UHDR_IMG_FMT_24bppYCbCr444:
1212 return getYuv444Pixel;
1213 case UHDR_IMG_FMT_16bppYCbCr422:
1214 return getYuv422Pixel;
1215 case UHDR_IMG_FMT_12bppYCbCr420:
1216 return getYuv420Pixel;
1217 case UHDR_IMG_FMT_24bppYCbCrP010:
1218 return getP010Pixel;
1219 case UHDR_IMG_FMT_30bppYCbCr444:
1220 return getYuv444Pixel10bit;
1221 case UHDR_IMG_FMT_32bppRGBA8888:
1222 return getRgba8888Pixel;
1223 case UHDR_IMG_FMT_32bppRGBA1010102:
1224 return getRgba1010102Pixel;
1225 case UHDR_IMG_FMT_64bppRGBAHalfFloat:
1226 return getRgbaF16Pixel;
1227 case UHDR_IMG_FMT_8bppYCbCr400:
1228 return getYuv400Pixel;
1229 case UHDR_IMG_FMT_24bppRGB888:
1230 return getRgb888Pixel;
1231 default:
1232 return nullptr;
1233 }
1234 return nullptr;
1235 }
1236
putPixelFn(uhdr_img_fmt_t format)1237 PutPixelFn putPixelFn(uhdr_img_fmt_t format) {
1238 switch (format) {
1239 case UHDR_IMG_FMT_24bppYCbCr444:
1240 return putYuv444Pixel;
1241 case UHDR_IMG_FMT_32bppRGBA8888:
1242 return putRgba8888Pixel;
1243 case UHDR_IMG_FMT_8bppYCbCr400:
1244 return putYuv400Pixel;
1245 case UHDR_IMG_FMT_24bppRGB888:
1246 return putRgb888Pixel;
1247 default:
1248 return nullptr;
1249 }
1250 return nullptr;
1251 }
1252
getSamplePixelFn(uhdr_img_fmt_t format)1253 SamplePixelFn getSamplePixelFn(uhdr_img_fmt_t format) {
1254 switch (format) {
1255 case UHDR_IMG_FMT_24bppYCbCr444:
1256 return sampleYuv444;
1257 case UHDR_IMG_FMT_16bppYCbCr422:
1258 return sampleYuv422;
1259 case UHDR_IMG_FMT_12bppYCbCr420:
1260 return sampleYuv420;
1261 case UHDR_IMG_FMT_24bppYCbCrP010:
1262 return sampleP010;
1263 case UHDR_IMG_FMT_30bppYCbCr444:
1264 return sampleYuv44410bit;
1265 case UHDR_IMG_FMT_32bppRGBA8888:
1266 return sampleRgba8888;
1267 case UHDR_IMG_FMT_32bppRGBA1010102:
1268 return sampleRgba1010102;
1269 case UHDR_IMG_FMT_64bppRGBAHalfFloat:
1270 return sampleRgbaF16;
1271 default:
1272 return nullptr;
1273 }
1274 return nullptr;
1275 }
1276
1277 ////////////////////////////////////////////////////////////////////////////////
1278 // common utils
1279
isPixelFormatRgb(uhdr_img_fmt_t format)1280 bool isPixelFormatRgb(uhdr_img_fmt_t format) {
1281 return format == UHDR_IMG_FMT_64bppRGBAHalfFloat || format == UHDR_IMG_FMT_32bppRGBA8888 ||
1282 format == UHDR_IMG_FMT_32bppRGBA1010102;
1283 }
1284
colorToRgba1010102(Color e_gamma)1285 uint32_t colorToRgba1010102(Color e_gamma) {
1286 uint32_t r = CLIP3((e_gamma.r * 1023 + 0.5f), 0.0f, 1023.0f);
1287 uint32_t g = CLIP3((e_gamma.g * 1023 + 0.5f), 0.0f, 1023.0f);
1288 uint32_t b = CLIP3((e_gamma.b * 1023 + 0.5f), 0.0f, 1023.0f);
1289 return (r | (g << 10) | (b << 20) | (0x3 << 30)); // Set alpha to 1.0
1290 }
1291
colorToRgbaF16(Color e_gamma)1292 uint64_t colorToRgbaF16(Color e_gamma) {
1293 return (uint64_t)floatToHalf(e_gamma.r) | (((uint64_t)floatToHalf(e_gamma.g)) << 16) |
1294 (((uint64_t)floatToHalf(e_gamma.b)) << 32) | (((uint64_t)floatToHalf(1.0f)) << 48);
1295 }
1296
convert_raw_input_to_ycbcr(uhdr_raw_image_t * src,bool chroma_sampling_enabled)1297 std::unique_ptr<uhdr_raw_image_ext_t> convert_raw_input_to_ycbcr(uhdr_raw_image_t* src,
1298 bool chroma_sampling_enabled) {
1299 std::unique_ptr<uhdr_raw_image_ext_t> dst = nullptr;
1300 Color (*rgbToyuv)(Color) = nullptr;
1301
1302 if (src->fmt == UHDR_IMG_FMT_32bppRGBA1010102 || src->fmt == UHDR_IMG_FMT_32bppRGBA8888) {
1303 if (src->cg == UHDR_CG_BT_709) {
1304 rgbToyuv = srgbRgbToYuv;
1305 } else if (src->cg == UHDR_CG_BT_2100) {
1306 rgbToyuv = bt2100RgbToYuv;
1307 } else if (src->cg == UHDR_CG_DISPLAY_P3) {
1308 rgbToyuv = p3RgbToYuv;
1309 } else {
1310 return dst;
1311 }
1312 }
1313
1314 if (src->fmt == UHDR_IMG_FMT_32bppRGBA1010102 && chroma_sampling_enabled) {
1315 dst = std::make_unique<uhdr_raw_image_ext_t>(UHDR_IMG_FMT_24bppYCbCrP010, src->cg, src->ct,
1316 UHDR_CR_FULL_RANGE, src->w, src->h, 64);
1317
1318 uint32_t* rgbData = static_cast<uint32_t*>(src->planes[UHDR_PLANE_PACKED]);
1319 unsigned int srcStride = src->stride[UHDR_PLANE_PACKED];
1320
1321 uint16_t* yData = static_cast<uint16_t*>(dst->planes[UHDR_PLANE_Y]);
1322 uint16_t* uData = static_cast<uint16_t*>(dst->planes[UHDR_PLANE_UV]);
1323 uint16_t* vData = uData + 1;
1324
1325 for (size_t i = 0; i < dst->h; i += 2) {
1326 for (size_t j = 0; j < dst->w; j += 2) {
1327 Color pixel[4];
1328
1329 pixel[0].r = float(rgbData[srcStride * i + j] & 0x3ff);
1330 pixel[0].g = float((rgbData[srcStride * i + j] >> 10) & 0x3ff);
1331 pixel[0].b = float((rgbData[srcStride * i + j] >> 20) & 0x3ff);
1332
1333 pixel[1].r = float(rgbData[srcStride * i + j + 1] & 0x3ff);
1334 pixel[1].g = float((rgbData[srcStride * i + j + 1] >> 10) & 0x3ff);
1335 pixel[1].b = float((rgbData[srcStride * i + j + 1] >> 20) & 0x3ff);
1336
1337 pixel[2].r = float(rgbData[srcStride * (i + 1) + j] & 0x3ff);
1338 pixel[2].g = float((rgbData[srcStride * (i + 1) + j] >> 10) & 0x3ff);
1339 pixel[2].b = float((rgbData[srcStride * (i + 1) + j] >> 20) & 0x3ff);
1340
1341 pixel[3].r = float(rgbData[srcStride * (i + 1) + j + 1] & 0x3ff);
1342 pixel[3].g = float((rgbData[srcStride * (i + 1) + j + 1] >> 10) & 0x3ff);
1343 pixel[3].b = float((rgbData[srcStride * (i + 1) + j + 1] >> 20) & 0x3ff);
1344
1345 for (int k = 0; k < 4; k++) {
1346 // Now we only support the RGB input being full range
1347 pixel[k] /= 1023.0f;
1348 pixel[k] = (*rgbToyuv)(pixel[k]);
1349
1350 pixel[k].y = (pixel[k].y * 1023.0f) + 0.5f;
1351 pixel[k].y = CLIP3(pixel[k].y, 0.0f, 1023.0f);
1352 }
1353
1354 yData[dst->stride[UHDR_PLANE_Y] * i + j] = uint16_t(pixel[0].y) << 6;
1355 yData[dst->stride[UHDR_PLANE_Y] * i + j + 1] = uint16_t(pixel[1].y) << 6;
1356 yData[dst->stride[UHDR_PLANE_Y] * (i + 1) + j] = uint16_t(pixel[2].y) << 6;
1357 yData[dst->stride[UHDR_PLANE_Y] * (i + 1) + j + 1] = uint16_t(pixel[3].y) << 6;
1358
1359 pixel[0].u = (pixel[0].u + pixel[1].u + pixel[2].u + pixel[3].u) / 4;
1360 pixel[0].v = (pixel[0].v + pixel[1].v + pixel[2].v + pixel[3].v) / 4;
1361
1362 pixel[0].u = (pixel[0].u * 1023.0f) + 512.0f + 0.5f;
1363 pixel[0].v = (pixel[0].v * 1023.0f) + 512.0f + 0.5f;
1364
1365 pixel[0].u = CLIP3(pixel[0].u, 0.0f, 1023.0f);
1366 pixel[0].v = CLIP3(pixel[0].v, 0.0f, 1023.0f);
1367
1368 uData[dst->stride[UHDR_PLANE_UV] * (i / 2) + j] = uint16_t(pixel[0].u) << 6;
1369 vData[dst->stride[UHDR_PLANE_UV] * (i / 2) + j] = uint16_t(pixel[0].v) << 6;
1370 }
1371 }
1372 } else if (src->fmt == UHDR_IMG_FMT_32bppRGBA1010102) {
1373 dst = std::make_unique<uhdr_raw_image_ext_t>(UHDR_IMG_FMT_30bppYCbCr444, src->cg, src->ct,
1374 UHDR_CR_FULL_RANGE, src->w, src->h, 64);
1375
1376 uint32_t* rgbData = static_cast<uint32_t*>(src->planes[UHDR_PLANE_PACKED]);
1377 unsigned int srcStride = src->stride[UHDR_PLANE_PACKED];
1378
1379 uint16_t* yData = static_cast<uint16_t*>(dst->planes[UHDR_PLANE_Y]);
1380 uint16_t* uData = static_cast<uint16_t*>(dst->planes[UHDR_PLANE_U]);
1381 uint16_t* vData = static_cast<uint16_t*>(dst->planes[UHDR_PLANE_V]);
1382
1383 for (size_t i = 0; i < dst->h; i++) {
1384 for (size_t j = 0; j < dst->w; j++) {
1385 Color pixel;
1386
1387 pixel.r = float(rgbData[srcStride * i + j] & 0x3ff);
1388 pixel.g = float((rgbData[srcStride * i + j] >> 10) & 0x3ff);
1389 pixel.b = float((rgbData[srcStride * i + j] >> 20) & 0x3ff);
1390
1391 // Now we only support the RGB input being full range
1392 pixel /= 1023.0f;
1393 pixel = (*rgbToyuv)(pixel);
1394
1395 pixel.y = (pixel.y * 1023.0f) + 0.5f;
1396 pixel.y = CLIP3(pixel.y, 0.0f, 1023.0f);
1397
1398 yData[dst->stride[UHDR_PLANE_Y] * i + j] = uint16_t(pixel.y);
1399
1400 pixel.u = (pixel.u * 1023.0f) + 512.0f + 0.5f;
1401 pixel.v = (pixel.v * 1023.0f) + 512.0f + 0.5f;
1402
1403 pixel.u = CLIP3(pixel.u, 0.0f, 1023.0f);
1404 pixel.v = CLIP3(pixel.v, 0.0f, 1023.0f);
1405
1406 uData[dst->stride[UHDR_PLANE_U] * i + j] = uint16_t(pixel.u);
1407 vData[dst->stride[UHDR_PLANE_V] * i + j] = uint16_t(pixel.v);
1408 }
1409 }
1410 } else if (src->fmt == UHDR_IMG_FMT_32bppRGBA8888 && chroma_sampling_enabled) {
1411 dst = std::make_unique<uhdr_raw_image_ext_t>(UHDR_IMG_FMT_12bppYCbCr420, src->cg, src->ct,
1412 UHDR_CR_FULL_RANGE, src->w, src->h, 64);
1413 uint32_t* rgbData = static_cast<uint32_t*>(src->planes[UHDR_PLANE_PACKED]);
1414 unsigned int srcStride = src->stride[UHDR_PLANE_PACKED];
1415
1416 uint8_t* yData = static_cast<uint8_t*>(dst->planes[UHDR_PLANE_Y]);
1417 uint8_t* uData = static_cast<uint8_t*>(dst->planes[UHDR_PLANE_U]);
1418 uint8_t* vData = static_cast<uint8_t*>(dst->planes[UHDR_PLANE_V]);
1419 for (size_t i = 0; i < dst->h; i += 2) {
1420 for (size_t j = 0; j < dst->w; j += 2) {
1421 Color pixel[4];
1422
1423 pixel[0].r = float(rgbData[srcStride * i + j] & 0xff);
1424 pixel[0].g = float((rgbData[srcStride * i + j] >> 8) & 0xff);
1425 pixel[0].b = float((rgbData[srcStride * i + j] >> 16) & 0xff);
1426
1427 pixel[1].r = float(rgbData[srcStride * i + (j + 1)] & 0xff);
1428 pixel[1].g = float((rgbData[srcStride * i + (j + 1)] >> 8) & 0xff);
1429 pixel[1].b = float((rgbData[srcStride * i + (j + 1)] >> 16) & 0xff);
1430
1431 pixel[2].r = float(rgbData[srcStride * (i + 1) + j] & 0xff);
1432 pixel[2].g = float((rgbData[srcStride * (i + 1) + j] >> 8) & 0xff);
1433 pixel[2].b = float((rgbData[srcStride * (i + 1) + j] >> 16) & 0xff);
1434
1435 pixel[3].r = float(rgbData[srcStride * (i + 1) + (j + 1)] & 0xff);
1436 pixel[3].g = float((rgbData[srcStride * (i + 1) + (j + 1)] >> 8) & 0xff);
1437 pixel[3].b = float((rgbData[srcStride * (i + 1) + (j + 1)] >> 16) & 0xff);
1438
1439 for (int k = 0; k < 4; k++) {
1440 // Now we only support the RGB input being full range
1441 pixel[k] /= 255.0f;
1442 pixel[k] = (*rgbToyuv)(pixel[k]);
1443
1444 pixel[k].y = pixel[k].y * 255.0f + 0.5f;
1445 pixel[k].y = CLIP3(pixel[k].y, 0.0f, 255.0f);
1446 }
1447 yData[dst->stride[UHDR_PLANE_Y] * i + j] = uint8_t(pixel[0].y);
1448 yData[dst->stride[UHDR_PLANE_Y] * i + j + 1] = uint8_t(pixel[1].y);
1449 yData[dst->stride[UHDR_PLANE_Y] * (i + 1) + j] = uint8_t(pixel[2].y);
1450 yData[dst->stride[UHDR_PLANE_Y] * (i + 1) + j + 1] = uint8_t(pixel[3].y);
1451
1452 pixel[0].u = (pixel[0].u + pixel[1].u + pixel[2].u + pixel[3].u) / 4;
1453 pixel[0].v = (pixel[0].v + pixel[1].v + pixel[2].v + pixel[3].v) / 4;
1454
1455 pixel[0].u = pixel[0].u * 255.0f + 0.5f + 128.0f;
1456 pixel[0].v = pixel[0].v * 255.0f + 0.5f + 128.0f;
1457
1458 pixel[0].u = CLIP3(pixel[0].u, 0.0f, 255.0f);
1459 pixel[0].v = CLIP3(pixel[0].v, 0.0f, 255.0f);
1460
1461 uData[dst->stride[UHDR_PLANE_U] * (i / 2) + (j / 2)] = uint8_t(pixel[0].u);
1462 vData[dst->stride[UHDR_PLANE_V] * (i / 2) + (j / 2)] = uint8_t(pixel[0].v);
1463 }
1464 }
1465 } else if (src->fmt == UHDR_IMG_FMT_32bppRGBA8888) {
1466 dst = std::make_unique<uhdr_raw_image_ext_t>(UHDR_IMG_FMT_24bppYCbCr444, src->cg, src->ct,
1467 UHDR_CR_FULL_RANGE, src->w, src->h, 64);
1468 uint32_t* rgbData = static_cast<uint32_t*>(src->planes[UHDR_PLANE_PACKED]);
1469 unsigned int srcStride = src->stride[UHDR_PLANE_PACKED];
1470
1471 uint8_t* yData = static_cast<uint8_t*>(dst->planes[UHDR_PLANE_Y]);
1472 uint8_t* uData = static_cast<uint8_t*>(dst->planes[UHDR_PLANE_U]);
1473 uint8_t* vData = static_cast<uint8_t*>(dst->planes[UHDR_PLANE_V]);
1474 for (size_t i = 0; i < dst->h; i++) {
1475 for (size_t j = 0; j < dst->w; j++) {
1476 Color pixel;
1477
1478 pixel.r = float(rgbData[srcStride * i + j] & 0xff);
1479 pixel.g = float((rgbData[srcStride * i + j] >> 8) & 0xff);
1480 pixel.b = float((rgbData[srcStride * i + j] >> 16) & 0xff);
1481
1482 // Now we only support the RGB input being full range
1483 pixel /= 255.0f;
1484 pixel = (*rgbToyuv)(pixel);
1485
1486 pixel.y = pixel.y * 255.0f + 0.5f;
1487 pixel.y = CLIP3(pixel.y, 0.0f, 255.0f);
1488 yData[dst->stride[UHDR_PLANE_Y] * i + j] = uint8_t(pixel.y);
1489
1490 pixel.u = pixel.u * 255.0f + 0.5f + 128.0f;
1491 pixel.v = pixel.v * 255.0f + 0.5f + 128.0f;
1492
1493 pixel.u = CLIP3(pixel.u, 0.0f, 255.0f);
1494 pixel.v = CLIP3(pixel.v, 0.0f, 255.0f);
1495
1496 uData[dst->stride[UHDR_PLANE_U] * i + j] = uint8_t(pixel.u);
1497 vData[dst->stride[UHDR_PLANE_V] * i + j] = uint8_t(pixel.v);
1498 }
1499 }
1500 } else if (src->fmt == UHDR_IMG_FMT_12bppYCbCr420 || src->fmt == UHDR_IMG_FMT_24bppYCbCrP010) {
1501 dst = std::make_unique<ultrahdr::uhdr_raw_image_ext_t>(src->fmt, src->cg, src->ct, src->range,
1502 src->w, src->h, 64);
1503 auto status = copy_raw_image(src, dst.get());
1504 if (status.error_code != UHDR_CODEC_OK) return nullptr;
1505 }
1506 return dst;
1507 }
1508
copy_raw_image(uhdr_raw_image_t * src)1509 std::unique_ptr<uhdr_raw_image_ext_t> copy_raw_image(uhdr_raw_image_t* src) {
1510 std::unique_ptr<uhdr_raw_image_ext_t> dst = std::make_unique<ultrahdr::uhdr_raw_image_ext_t>(
1511 src->fmt, src->cg, src->ct, src->range, src->w, src->h, 64);
1512 auto status = copy_raw_image(src, dst.get());
1513 if (status.error_code != UHDR_CODEC_OK) return nullptr;
1514 return dst;
1515 }
1516
copy_raw_image(uhdr_raw_image_t * src,uhdr_raw_image_t * dst)1517 uhdr_error_info_t copy_raw_image(uhdr_raw_image_t* src, uhdr_raw_image_t* dst) {
1518 if (dst->w != src->w || dst->h != src->h) {
1519 uhdr_error_info_t status;
1520 status.error_code = UHDR_CODEC_MEM_ERROR;
1521 status.has_detail = 1;
1522 snprintf(status.detail, sizeof status.detail,
1523 "destination image dimensions %dx%d and source image dimensions %dx%d are not "
1524 "identical for copy_raw_image",
1525 dst->w, dst->h, src->w, src->h);
1526 return status;
1527 }
1528
1529 dst->cg = src->cg;
1530 dst->ct = src->ct;
1531 dst->range = src->range;
1532 if (dst->fmt == src->fmt) {
1533 if (src->fmt == UHDR_IMG_FMT_24bppYCbCrP010) {
1534 size_t bpp = 2;
1535 uint8_t* y_dst = static_cast<uint8_t*>(dst->planes[UHDR_PLANE_Y]);
1536 uint8_t* y_src = static_cast<uint8_t*>(src->planes[UHDR_PLANE_Y]);
1537 uint8_t* uv_dst = static_cast<uint8_t*>(dst->planes[UHDR_PLANE_UV]);
1538 uint8_t* uv_src = static_cast<uint8_t*>(src->planes[UHDR_PLANE_UV]);
1539
1540 // copy y
1541 for (size_t i = 0; i < src->h; i++) {
1542 memcpy(y_dst, y_src, src->w * bpp);
1543 y_dst += (dst->stride[UHDR_PLANE_Y] * bpp);
1544 y_src += (src->stride[UHDR_PLANE_Y] * bpp);
1545 }
1546 // copy cbcr
1547 for (size_t i = 0; i < src->h / 2; i++) {
1548 memcpy(uv_dst, uv_src, src->w * bpp);
1549 uv_dst += (dst->stride[UHDR_PLANE_UV] * bpp);
1550 uv_src += (src->stride[UHDR_PLANE_UV] * bpp);
1551 }
1552 return g_no_error;
1553 } else if (src->fmt == UHDR_IMG_FMT_12bppYCbCr420) {
1554 uint8_t* y_dst = static_cast<uint8_t*>(dst->planes[UHDR_PLANE_Y]);
1555 uint8_t* y_src = static_cast<uint8_t*>(src->planes[UHDR_PLANE_Y]);
1556 uint8_t* u_dst = static_cast<uint8_t*>(dst->planes[UHDR_PLANE_U]);
1557 uint8_t* u_src = static_cast<uint8_t*>(src->planes[UHDR_PLANE_U]);
1558 uint8_t* v_dst = static_cast<uint8_t*>(dst->planes[UHDR_PLANE_V]);
1559 uint8_t* v_src = static_cast<uint8_t*>(src->planes[UHDR_PLANE_V]);
1560
1561 // copy y
1562 for (size_t i = 0; i < src->h; i++) {
1563 memcpy(y_dst, y_src, src->w);
1564 y_dst += dst->stride[UHDR_PLANE_Y];
1565 y_src += src->stride[UHDR_PLANE_Y];
1566 }
1567 // copy cb & cr
1568 for (size_t i = 0; i < src->h / 2; i++) {
1569 memcpy(u_dst, u_src, src->w / 2);
1570 memcpy(v_dst, v_src, src->w / 2);
1571 u_dst += dst->stride[UHDR_PLANE_U];
1572 v_dst += dst->stride[UHDR_PLANE_V];
1573 u_src += src->stride[UHDR_PLANE_U];
1574 v_src += src->stride[UHDR_PLANE_V];
1575 }
1576 return g_no_error;
1577 } else if (src->fmt == UHDR_IMG_FMT_8bppYCbCr400 || src->fmt == UHDR_IMG_FMT_32bppRGBA8888 ||
1578 src->fmt == UHDR_IMG_FMT_64bppRGBAHalfFloat ||
1579 src->fmt == UHDR_IMG_FMT_32bppRGBA1010102 || src->fmt == UHDR_IMG_FMT_24bppRGB888) {
1580 uint8_t* plane_dst = static_cast<uint8_t*>(dst->planes[UHDR_PLANE_PACKED]);
1581 uint8_t* plane_src = static_cast<uint8_t*>(src->planes[UHDR_PLANE_PACKED]);
1582 size_t bpp = 1;
1583
1584 if (src->fmt == UHDR_IMG_FMT_32bppRGBA1010102 || src->fmt == UHDR_IMG_FMT_32bppRGBA8888)
1585 bpp = 4;
1586 else if (src->fmt == UHDR_IMG_FMT_64bppRGBAHalfFloat)
1587 bpp = 8;
1588 else if (src->fmt == UHDR_IMG_FMT_24bppRGB888)
1589 bpp = 3;
1590 for (size_t i = 0; i < src->h; i++) {
1591 memcpy(plane_dst, plane_src, src->w * bpp);
1592 plane_dst += (bpp * dst->stride[UHDR_PLANE_PACKED]);
1593 plane_src += (bpp * src->stride[UHDR_PLANE_PACKED]);
1594 }
1595 return g_no_error;
1596 }
1597 } else {
1598 if (src->fmt == UHDR_IMG_FMT_24bppRGB888 && dst->fmt == UHDR_IMG_FMT_32bppRGBA8888) {
1599 uint32_t* plane_dst = static_cast<uint32_t*>(dst->planes[UHDR_PLANE_PACKED]);
1600 uint8_t* plane_src = static_cast<uint8_t*>(src->planes[UHDR_PLANE_PACKED]);
1601 for (size_t i = 0; i < src->h; i++) {
1602 uint32_t* pixel_dst = plane_dst;
1603 uint8_t* pixel_src = plane_src;
1604 for (size_t j = 0; j < src->w; j++) {
1605 *pixel_dst = pixel_src[0] | (pixel_src[1] << 8) | (pixel_src[2] << 16) | (0xff << 24);
1606 pixel_src += 3;
1607 pixel_dst += 1;
1608 }
1609 plane_dst += dst->stride[UHDR_PLANE_PACKED];
1610 plane_src += (size_t)3 * src->stride[UHDR_PLANE_PACKED];
1611 }
1612 return g_no_error;
1613 }
1614 }
1615 uhdr_error_info_t status;
1616 status.error_code = UHDR_CODEC_UNSUPPORTED_FEATURE;
1617 status.has_detail = 1;
1618 snprintf(
1619 status.detail, sizeof status.detail,
1620 "unsupported source / destinations color formats in copy_raw_image, src fmt %d, dst fmt %d",
1621 src->fmt, dst->fmt);
1622 return status;
1623 }
1624
1625 // Use double type for intermediate results for better precision.
floatToUnsignedFractionImpl(float v,uint32_t maxNumerator,uint32_t * numerator,uint32_t * denominator)1626 static bool floatToUnsignedFractionImpl(float v, uint32_t maxNumerator, uint32_t* numerator,
1627 uint32_t* denominator) {
1628 if (std::isnan(v) || v < 0 || v > maxNumerator) {
1629 return false;
1630 }
1631
1632 // Maximum denominator: makes sure that the numerator is <= maxNumerator and the denominator
1633 // is <= UINT32_MAX.
1634 const uint64_t maxD = (v <= 1) ? UINT32_MAX : (uint64_t)floor(maxNumerator / v);
1635
1636 // Find the best approximation of v as a fraction using continued fractions, see
1637 // https://en.wikipedia.org/wiki/Continued_fraction
1638 *denominator = 1;
1639 uint32_t previousD = 0;
1640 double currentV = (double)v - floor(v);
1641 int iter = 0;
1642 // Set a maximum number of iterations to be safe. Most numbers should
1643 // converge in less than ~20 iterations.
1644 // The golden ratio is the worst case and takes 39 iterations.
1645 const int maxIter = 39;
1646 while (iter < maxIter) {
1647 const double numeratorDouble = (double)(*denominator) * v;
1648 if (numeratorDouble > maxNumerator) {
1649 return false;
1650 }
1651 *numerator = (uint32_t)round(numeratorDouble);
1652 if (fabs(numeratorDouble - (*numerator)) == 0.0) {
1653 return true;
1654 }
1655 currentV = 1.0 / currentV;
1656 const double newD = previousD + floor(currentV) * (*denominator);
1657 if (newD > maxD) {
1658 // This is the best we can do with a denominator <= max_d.
1659 return true;
1660 }
1661 previousD = *denominator;
1662 if (newD > (double)UINT32_MAX) {
1663 return false;
1664 }
1665 *denominator = (uint32_t)newD;
1666 currentV -= floor(currentV);
1667 ++iter;
1668 }
1669 // Maximum number of iterations reached, return what we've found.
1670 // For max_iter >= 39 we shouldn't get here. max_iter can be set
1671 // to a lower value to speed up the algorithm if needed.
1672 *numerator = (uint32_t)round((double)(*denominator) * v);
1673 return true;
1674 }
1675
floatToSignedFraction(float v,int32_t * numerator,uint32_t * denominator)1676 bool floatToSignedFraction(float v, int32_t* numerator, uint32_t* denominator) {
1677 uint32_t positive_numerator;
1678 if (!floatToUnsignedFractionImpl(fabs(v), INT32_MAX, &positive_numerator, denominator)) {
1679 return false;
1680 }
1681 *numerator = (int32_t)positive_numerator;
1682 if (v < 0) {
1683 *numerator *= -1;
1684 }
1685 return true;
1686 }
1687
floatToUnsignedFraction(float v,uint32_t * numerator,uint32_t * denominator)1688 bool floatToUnsignedFraction(float v, uint32_t* numerator, uint32_t* denominator) {
1689 return floatToUnsignedFractionImpl(v, UINT32_MAX, numerator, denominator);
1690 }
1691
1692 } // namespace ultrahdr
1693