1 /*
2 * Copyright (c) 2020, Alliance for Open Media. All rights reserved
3 *
4 * This source code is subject to the terms of the BSD 2 Clause License and
5 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
6 * was not distributed with this source code in the LICENSE file, you can
7 * obtain it at www.aomedia.org/license/software. If the Alliance for Open
8 * Media Patent License 1.0 was not distributed with this source code in the
9 * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
10 */
11
12 #include <assert.h>
13 #include <limits.h>
14 #include <math.h>
15
16 #include "config/aom_dsp_rtcd.h"
17 #include "aom_dsp/aom_dsp_common.h"
18 #include "aom_scale/yv12config.h"
19 #include "aom/aom_integer.h"
20 #include "av1/encoder/context_tree.h"
21 #include "av1/encoder/av1_noise_estimate.h"
22 #include "av1/encoder/encoder.h"
23 #if CONFIG_AV1_TEMPORAL_DENOISING
24 #include "av1/encoder/av1_temporal_denoiser.h"
25 #endif
26
27 #if CONFIG_AV1_TEMPORAL_DENOISING
28 // For SVC: only do noise estimation on top spatial layer.
noise_est_svc(const struct AV1_COMP * const cpi)29 static INLINE int noise_est_svc(const struct AV1_COMP *const cpi) {
30 return (!cpi->ppi->use_svc ||
31 (cpi->ppi->use_svc &&
32 cpi->svc.spatial_layer_id == cpi->svc.number_spatial_layers - 1));
33 }
34 #endif
35
av1_noise_estimate_init(NOISE_ESTIMATE * const ne,int width,int height)36 void av1_noise_estimate_init(NOISE_ESTIMATE *const ne, int width, int height) {
37 ne->enabled = 0;
38 ne->level = (width * height < 1280 * 720) ? kLowLow : kLow;
39 ne->value = 0;
40 ne->count = 0;
41 ne->thresh = 90;
42 ne->last_w = 0;
43 ne->last_h = 0;
44 if (width * height >= 1920 * 1080) {
45 ne->thresh = 200;
46 } else if (width * height >= 1280 * 720) {
47 ne->thresh = 140;
48 } else if (width * height >= 640 * 360) {
49 ne->thresh = 115;
50 }
51 ne->num_frames_estimate = 15;
52 ne->adapt_thresh = (3 * ne->thresh) >> 1;
53 }
54
enable_noise_estimation(AV1_COMP * const cpi)55 static int enable_noise_estimation(AV1_COMP *const cpi) {
56 const int resize_pending = is_frame_resize_pending(cpi);
57
58 #if CONFIG_AV1_HIGHBITDEPTH
59 if (cpi->common.seq_params->use_highbitdepth) return 0;
60 #endif
61 // Enable noise estimation if denoising is on.
62 #if CONFIG_AV1_TEMPORAL_DENOISING
63 if (cpi->oxcf.noise_sensitivity > 0 && noise_est_svc(cpi) &&
64 cpi->common.width >= 320 && cpi->common.height >= 180)
65 return 1;
66 #endif
67 // Only allow noise estimate under certain encoding mode.
68 // Enabled for 1 pass CBR, speed >=5, and if resolution is same as original.
69 // Not enabled for SVC mode and screen_content_mode.
70 // Not enabled for low resolutions.
71 if (cpi->oxcf.pass == AOM_RC_ONE_PASS && cpi->oxcf.rc_cfg.mode == AOM_CBR &&
72 cpi->oxcf.q_cfg.aq_mode == CYCLIC_REFRESH_AQ && cpi->oxcf.speed >= 5 &&
73 resize_pending == 0 && !cpi->ppi->use_svc &&
74 cpi->oxcf.tune_cfg.content != AOM_CONTENT_SCREEN &&
75 cpi->common.width * cpi->common.height >= 640 * 360)
76 return 1;
77 else
78 return 0;
79 }
80
81 #if CONFIG_AV1_TEMPORAL_DENOISING
copy_frame(YV12_BUFFER_CONFIG * const dest,const YV12_BUFFER_CONFIG * const src)82 static void copy_frame(YV12_BUFFER_CONFIG *const dest,
83 const YV12_BUFFER_CONFIG *const src) {
84 const uint8_t *srcbuf = src->y_buffer;
85 uint8_t *destbuf = dest->y_buffer;
86
87 assert(dest->y_width == src->y_width);
88 assert(dest->y_height == src->y_height);
89
90 for (int r = 0; r < dest->y_height; ++r) {
91 memcpy(destbuf, srcbuf, dest->y_width);
92 destbuf += dest->y_stride;
93 srcbuf += src->y_stride;
94 }
95 }
96 #endif // CONFIG_AV1_TEMPORAL_DENOISING
97
av1_noise_estimate_extract_level(NOISE_ESTIMATE * const ne)98 NOISE_LEVEL av1_noise_estimate_extract_level(NOISE_ESTIMATE *const ne) {
99 int noise_level = kLowLow;
100 if (ne->value > (ne->thresh << 1)) {
101 noise_level = kHigh;
102 } else {
103 if (ne->value > ne->thresh)
104 noise_level = kMedium;
105 else if (ne->value > (ne->thresh >> 1))
106 noise_level = kLow;
107 else
108 noise_level = kLowLow;
109 }
110 return noise_level;
111 }
112
av1_update_noise_estimate(AV1_COMP * const cpi)113 void av1_update_noise_estimate(AV1_COMP *const cpi) {
114 const AV1_COMMON *const cm = &cpi->common;
115 const CommonModeInfoParams *const mi_params = &cm->mi_params;
116
117 NOISE_ESTIMATE *const ne = &cpi->noise_estimate;
118 const int low_res = (cm->width <= 352 && cm->height <= 288);
119 // Estimate of noise level every frame_period frames.
120 int frame_period = 8;
121 int thresh_consec_zeromv = 2;
122 int frame_counter = cm->current_frame.frame_number;
123 // Estimate is between current source and last source.
124 YV12_BUFFER_CONFIG *last_source = cpi->last_source;
125 #if CONFIG_AV1_TEMPORAL_DENOISING
126 if (cpi->oxcf.noise_sensitivity > 0 && noise_est_svc(cpi)) {
127 last_source = &cpi->denoiser.last_source;
128 // Tune these thresholds for different resolutions when denoising is
129 // enabled.
130 if (cm->width > 640 && cm->width <= 1920) {
131 thresh_consec_zeromv = 2;
132 }
133 }
134 #endif
135 ne->enabled = enable_noise_estimation(cpi);
136 if (cpi->svc.number_spatial_layers > 1)
137 frame_counter = cpi->svc.current_superframe;
138 if (!ne->enabled || frame_counter % frame_period != 0 ||
139 last_source == NULL ||
140 (cpi->svc.number_spatial_layers == 1 &&
141 (ne->last_w != cm->width || ne->last_h != cm->height))) {
142 #if CONFIG_AV1_TEMPORAL_DENOISING
143 if (cpi->oxcf.noise_sensitivity > 0 && noise_est_svc(cpi))
144 copy_frame(&cpi->denoiser.last_source, cpi->source);
145 #endif
146 if (last_source != NULL) {
147 ne->last_w = cm->width;
148 ne->last_h = cm->height;
149 }
150 return;
151 } else if (frame_counter > 60 && cpi->svc.num_encoded_top_layer > 1 &&
152 cpi->rc.frames_since_key > cpi->svc.number_spatial_layers &&
153 cpi->svc.spatial_layer_id == cpi->svc.number_spatial_layers - 1 &&
154 cpi->rc.avg_frame_low_motion < (low_res ? 60 : 40)) {
155 // Force noise estimation to 0 and denoiser off if content has high motion.
156 ne->level = kLowLow;
157 ne->count = 0;
158 ne->num_frames_estimate = 10;
159 #if CONFIG_AV1_TEMPORAL_DENOISING
160 if (cpi->oxcf.noise_sensitivity > 0 && noise_est_svc(cpi) &&
161 cpi->svc.current_superframe > 1) {
162 av1_denoiser_set_noise_level(cpi, ne->level);
163 copy_frame(&cpi->denoiser.last_source, cpi->source);
164 }
165 #endif
166 return;
167 } else {
168 unsigned int bin_size = 100;
169 unsigned int hist[MAX_VAR_HIST_BINS] = { 0 };
170 unsigned int hist_avg[MAX_VAR_HIST_BINS];
171 unsigned int max_bin = 0;
172 unsigned int max_bin_count = 0;
173 unsigned int bin_cnt;
174 int bsize = BLOCK_16X16;
175 // Loop over sub-sample of 16x16 blocks of frame, and for blocks that have
176 // been encoded as zero/small mv at least x consecutive frames, compute
177 // the variance to update estimate of noise in the source.
178 const uint8_t *src_y = cpi->source->y_buffer;
179 const int src_ystride = cpi->source->y_stride;
180 const uint8_t *last_src_y = last_source->y_buffer;
181 const int last_src_ystride = last_source->y_stride;
182 int mi_row, mi_col;
183 int num_low_motion = 0;
184 int frame_low_motion = 1;
185 for (mi_row = 0; mi_row < mi_params->mi_rows; mi_row += 2) {
186 for (mi_col = 0; mi_col < mi_params->mi_cols; mi_col += 2) {
187 int bl_index =
188 (mi_row >> 1) * (mi_params->mi_cols >> 1) + (mi_col >> 1);
189 if (cpi->consec_zero_mv[bl_index] > thresh_consec_zeromv)
190 num_low_motion++;
191 }
192 }
193 if (num_low_motion <
194 (((3 * (mi_params->mi_rows * mi_params->mi_cols) >> 2)) >> 3))
195 frame_low_motion = 0;
196 for (mi_row = 0; mi_row < mi_params->mi_rows; mi_row++) {
197 for (mi_col = 0; mi_col < mi_params->mi_cols; mi_col++) {
198 // 16x16 blocks, 1/4 sample of frame.
199 if (mi_row % 8 == 0 && mi_col % 8 == 0 &&
200 mi_row < mi_params->mi_rows - 3 &&
201 mi_col < mi_params->mi_cols - 3) {
202 int bl_index =
203 (mi_row >> 1) * (mi_params->mi_cols >> 1) + (mi_col >> 1);
204 int bl_index1 = bl_index + 1;
205 int bl_index2 = bl_index + (mi_params->mi_cols >> 1);
206 int bl_index3 = bl_index2 + 1;
207 int consec_zeromv =
208 AOMMIN(cpi->consec_zero_mv[bl_index],
209 AOMMIN(cpi->consec_zero_mv[bl_index1],
210 AOMMIN(cpi->consec_zero_mv[bl_index2],
211 cpi->consec_zero_mv[bl_index3])));
212 // Only consider blocks that are likely steady background. i.e, have
213 // been encoded as zero/low motion x (= thresh_consec_zeromv) frames
214 // in a row. consec_zero_mv[] defined for 8x8 blocks, so consider all
215 // 4 sub-blocks for 16x16 block. And exclude this frame if
216 // high_source_sad is true (i.e., scene/content change).
217 if (frame_low_motion && consec_zeromv > thresh_consec_zeromv &&
218 !cpi->rc.high_source_sad) {
219 unsigned int sse;
220 // Compute variance between co-located blocks from current and
221 // last input frames.
222 unsigned int variance = cpi->ppi->fn_ptr[bsize].vf(
223 src_y, src_ystride, last_src_y, last_src_ystride, &sse);
224 unsigned int hist_index = variance / bin_size;
225 if (hist_index < MAX_VAR_HIST_BINS)
226 hist[hist_index]++;
227 else if (hist_index < 3 * (MAX_VAR_HIST_BINS >> 1))
228 hist[MAX_VAR_HIST_BINS - 1]++; // Account for the tail
229 }
230 }
231 src_y += 4;
232 last_src_y += 4;
233 }
234 src_y += (src_ystride << 2) - (mi_params->mi_cols << 2);
235 last_src_y += (last_src_ystride << 2) - (mi_params->mi_cols << 2);
236 }
237 ne->last_w = cm->width;
238 ne->last_h = cm->height;
239 // Adjust histogram to account for effect that histogram flattens
240 // and shifts to zero as scene darkens.
241 if (hist[0] > 10 && (hist[MAX_VAR_HIST_BINS - 1] > hist[0] >> 2)) {
242 hist[0] = 0;
243 hist[1] >>= 2;
244 hist[2] >>= 2;
245 hist[3] >>= 2;
246 hist[4] >>= 1;
247 hist[5] >>= 1;
248 hist[6] = 3 * hist[6] >> 1;
249 hist[MAX_VAR_HIST_BINS - 1] >>= 1;
250 }
251
252 // Average hist[] and find largest bin
253 for (bin_cnt = 0; bin_cnt < MAX_VAR_HIST_BINS; bin_cnt++) {
254 if (bin_cnt == 0)
255 hist_avg[bin_cnt] = (hist[0] + hist[1] + hist[2]) / 3;
256 else if (bin_cnt == MAX_VAR_HIST_BINS - 1)
257 hist_avg[bin_cnt] = hist[MAX_VAR_HIST_BINS - 1] >> 2;
258 else if (bin_cnt == MAX_VAR_HIST_BINS - 2)
259 hist_avg[bin_cnt] = (hist[bin_cnt - 1] + 2 * hist[bin_cnt] +
260 (hist[bin_cnt + 1] >> 1) + 2) >>
261 2;
262 else
263 hist_avg[bin_cnt] =
264 (hist[bin_cnt - 1] + 2 * hist[bin_cnt] + hist[bin_cnt + 1] + 2) >>
265 2;
266
267 if (hist_avg[bin_cnt] > max_bin_count) {
268 max_bin_count = hist_avg[bin_cnt];
269 max_bin = bin_cnt;
270 }
271 }
272 // Scale by 40 to work with existing thresholds
273 ne->value = (int)((3 * ne->value + max_bin * 40) >> 2);
274 // Quickly increase VNR strength when the noise level increases suddenly.
275 if (ne->level < kMedium && ne->value > ne->adapt_thresh) {
276 ne->count = ne->num_frames_estimate;
277 } else {
278 ne->count++;
279 }
280 if (ne->count == ne->num_frames_estimate) {
281 // Reset counter and check noise level condition.
282 ne->num_frames_estimate = 30;
283 ne->count = 0;
284 ne->level = av1_noise_estimate_extract_level(ne);
285 #if CONFIG_AV1_TEMPORAL_DENOISING
286 if (cpi->oxcf.noise_sensitivity > 0 && noise_est_svc(cpi))
287 av1_denoiser_set_noise_level(cpi, ne->level);
288 #endif
289 }
290 }
291 #if CONFIG_AV1_TEMPORAL_DENOISING
292 if (cpi->oxcf.noise_sensitivity > 0 && noise_est_svc(cpi))
293 copy_frame(&cpi->denoiser.last_source, cpi->source);
294 #endif
295 }
296