1 /*
2 * Copyright (c) 2020 Paul B Mahol
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
10 *
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21 /**
22 * @file
23 * Cineform HD video encoder
24 */
25
26 #include <stdlib.h>
27 #include <string.h>
28
29 #include "libavutil/avassert.h"
30 #include "libavutil/imgutils.h"
31 #include "libavutil/opt.h"
32
33 #include "avcodec.h"
34 #include "bytestream.h"
35 #include "cfhd.h"
36 #include "cfhdencdsp.h"
37 #include "put_bits.h"
38 #include "internal.h"
39 #include "thread.h"
40
41 /* Derived from existing tables from decoder */
42 static const unsigned codebook[256][2] = {
43 { 1, 0x00000000 }, { 2, 0x00000002 }, { 3, 0x00000007 }, { 5, 0x00000019 }, { 6, 0x00000030 },
44 { 6, 0x00000036 }, { 7, 0x00000063 }, { 7, 0x0000006B }, { 7, 0x0000006F }, { 8, 0x000000D4 },
45 { 8, 0x000000DC }, { 9, 0x00000189 }, { 9, 0x000001A0 }, { 9, 0x000001AB }, {10, 0x00000310 },
46 {10, 0x00000316 }, {10, 0x00000354 }, {10, 0x00000375 }, {10, 0x00000377 }, {11, 0x00000623 },
47 {11, 0x00000684 }, {11, 0x000006AB }, {11, 0x000006EC }, {12, 0x00000C44 }, {12, 0x00000C5C },
48 {12, 0x00000C5E }, {12, 0x00000D55 }, {12, 0x00000DD1 }, {12, 0x00000DD3 }, {12, 0x00000DDB },
49 {13, 0x0000188B }, {13, 0x000018BB }, {13, 0x00001AA8 }, {13, 0x00001BA0 }, {13, 0x00001BA4 },
50 {13, 0x00001BB5 }, {14, 0x00003115 }, {14, 0x00003175 }, {14, 0x0000317D }, {14, 0x00003553 },
51 {14, 0x00003768 }, {15, 0x00006228 }, {15, 0x000062E8 }, {15, 0x000062F8 }, {15, 0x00006AA4 },
52 {15, 0x00006E85 }, {15, 0x00006E87 }, {15, 0x00006ED3 }, {16, 0x0000C453 }, {16, 0x0000C5D3 },
53 {16, 0x0000C5F3 }, {16, 0x0000DD08 }, {16, 0x0000DD0C }, {16, 0x0000DDA4 }, {17, 0x000188A4 },
54 {17, 0x00018BA5 }, {17, 0x00018BE5 }, {17, 0x0001AA95 }, {17, 0x0001AA97 }, {17, 0x0001BA13 },
55 {17, 0x0001BB4A }, {17, 0x0001BB4B }, {18, 0x00031748 }, {18, 0x000317C8 }, {18, 0x00035528 },
56 {18, 0x0003552C }, {18, 0x00037424 }, {18, 0x00037434 }, {18, 0x00037436 }, {19, 0x00062294 },
57 {19, 0x00062E92 }, {19, 0x00062F92 }, {19, 0x0006AA52 }, {19, 0x0006AA5A }, {19, 0x0006E84A },
58 {19, 0x0006E86A }, {19, 0x0006E86E }, {20, 0x000C452A }, {20, 0x000C5D27 }, {20, 0x000C5F26 },
59 {20, 0x000D54A6 }, {20, 0x000D54B6 }, {20, 0x000DD096 }, {20, 0x000DD0D6 }, {20, 0x000DD0DE },
60 {21, 0x00188A56 }, {21, 0x0018BA4D }, {21, 0x0018BE4E }, {21, 0x0018BE4F }, {21, 0x001AA96E },
61 {21, 0x001BA12E }, {21, 0x001BA12F }, {21, 0x001BA1AF }, {21, 0x001BA1BF }, {22, 0x00317498 },
62 {22, 0x0035529C }, {22, 0x0035529D }, {22, 0x003552DE }, {22, 0x003552DF }, {22, 0x0037435D },
63 {22, 0x0037437D }, {23, 0x0062295D }, {23, 0x0062E933 }, {23, 0x006AA53D }, {23, 0x006AA53E },
64 {23, 0x006AA53F }, {23, 0x006E86B9 }, {23, 0x006E86F8 }, {24, 0x00C452B8 }, {24, 0x00C5D265 },
65 {24, 0x00D54A78 }, {24, 0x00D54A79 }, {24, 0x00DD0D70 }, {24, 0x00DD0D71 }, {24, 0x00DD0DF2 },
66 {24, 0x00DD0DF3 }, {26, 0x03114BA2 }, {25, 0x0188A5B1 }, {25, 0x0188A58B }, {25, 0x0188A595 },
67 {25, 0x0188A5D6 }, {25, 0x0188A5D7 }, {25, 0x0188A5A8 }, {25, 0x0188A5AE }, {25, 0x0188A5AF },
68 {25, 0x0188A5C4 }, {25, 0x0188A5C5 }, {25, 0x0188A587 }, {25, 0x0188A584 }, {25, 0x0188A585 },
69 {25, 0x0188A5C6 }, {25, 0x0188A5C7 }, {25, 0x0188A5CC }, {25, 0x0188A5CD }, {25, 0x0188A581 },
70 {25, 0x0188A582 }, {25, 0x0188A583 }, {25, 0x0188A5CE }, {25, 0x0188A5CF }, {25, 0x0188A5C2 },
71 {25, 0x0188A5C3 }, {25, 0x0188A5C1 }, {25, 0x0188A5B4 }, {25, 0x0188A5B5 }, {25, 0x0188A5E6 },
72 {25, 0x0188A5E7 }, {25, 0x0188A5E4 }, {25, 0x0188A5E5 }, {25, 0x0188A5AB }, {25, 0x0188A5E0 },
73 {25, 0x0188A5E1 }, {25, 0x0188A5E2 }, {25, 0x0188A5E3 }, {25, 0x0188A5B6 }, {25, 0x0188A5B7 },
74 {25, 0x0188A5FD }, {25, 0x0188A57E }, {25, 0x0188A57F }, {25, 0x0188A5EC }, {25, 0x0188A5ED },
75 {25, 0x0188A5FE }, {25, 0x0188A5FF }, {25, 0x0188A57D }, {25, 0x0188A59C }, {25, 0x0188A59D },
76 {25, 0x0188A5E8 }, {25, 0x0188A5E9 }, {25, 0x0188A5EA }, {25, 0x0188A5EB }, {25, 0x0188A5EF },
77 {25, 0x0188A57A }, {25, 0x0188A57B }, {25, 0x0188A578 }, {25, 0x0188A579 }, {25, 0x0188A5BA },
78 {25, 0x0188A5BB }, {25, 0x0188A5B8 }, {25, 0x0188A5B9 }, {25, 0x0188A588 }, {25, 0x0188A589 },
79 {25, 0x018BA4C8 }, {25, 0x018BA4C9 }, {25, 0x0188A5FA }, {25, 0x0188A5FB }, {25, 0x0188A5BC },
80 {25, 0x0188A5BD }, {25, 0x0188A598 }, {25, 0x0188A599 }, {25, 0x0188A5F4 }, {25, 0x0188A5F5 },
81 {25, 0x0188A59B }, {25, 0x0188A5DE }, {25, 0x0188A5DF }, {25, 0x0188A596 }, {25, 0x0188A597 },
82 {25, 0x0188A5F8 }, {25, 0x0188A5F9 }, {25, 0x0188A5F1 }, {25, 0x0188A58E }, {25, 0x0188A58F },
83 {25, 0x0188A5DC }, {25, 0x0188A5DD }, {25, 0x0188A5F2 }, {25, 0x0188A5F3 }, {25, 0x0188A58C },
84 {25, 0x0188A58D }, {25, 0x0188A5A4 }, {25, 0x0188A5F0 }, {25, 0x0188A5A5 }, {25, 0x0188A5A6 },
85 {25, 0x0188A5A7 }, {25, 0x0188A59A }, {25, 0x0188A5A2 }, {25, 0x0188A5A3 }, {25, 0x0188A58A },
86 {25, 0x0188A5B0 }, {25, 0x0188A5A0 }, {25, 0x0188A5A1 }, {25, 0x0188A5DA }, {25, 0x0188A5DB },
87 {25, 0x0188A59E }, {25, 0x0188A59F }, {25, 0x0188A5D8 }, {25, 0x0188A5EE }, {25, 0x0188A5D9 },
88 {25, 0x0188A5F6 }, {25, 0x0188A5F7 }, {25, 0x0188A57C }, {25, 0x0188A5C8 }, {25, 0x0188A5C9 },
89 {25, 0x0188A594 }, {25, 0x0188A5FC }, {25, 0x0188A5CA }, {25, 0x0188A5CB }, {25, 0x0188A5B2 },
90 {25, 0x0188A5AA }, {25, 0x0188A5B3 }, {25, 0x0188A572 }, {25, 0x0188A573 }, {25, 0x0188A5C0 },
91 {25, 0x0188A5BE }, {25, 0x0188A5BF }, {25, 0x0188A592 }, {25, 0x0188A580 }, {25, 0x0188A593 },
92 {25, 0x0188A590 }, {25, 0x0188A591 }, {25, 0x0188A586 }, {25, 0x0188A5A9 }, {25, 0x0188A5D2 },
93 {25, 0x0188A5D3 }, {25, 0x0188A5D4 }, {25, 0x0188A5D5 }, {25, 0x0188A5AC }, {25, 0x0188A5AD },
94 {25, 0x0188A5D0 },
95 };
96
97 /* Derived by extracting runcodes from existing tables from decoder */
98 static const uint16_t runbook[18][3] = {
99 {1, 0x0000, 1}, {2, 0x0000, 2}, {3, 0x0000, 3}, {4, 0x0000, 4},
100 {5, 0x0000, 5}, {6, 0x0000, 6}, {7, 0x0000, 7}, {8, 0x0000, 8},
101 {9, 0x0000, 9}, {10, 0x0000, 10}, {11, 0x0000, 11},
102 {7, 0x0069, 12}, {8, 0x00D1, 20}, {9, 0x018A, 32},
103 {10, 0x0343, 60}, {11, 0x0685, 100}, {13, 0x18BF, 180}, {13, 0x1BA5, 320},
104 };
105
106 /*
107 * Derived by inspecting various quality encodes
108 * and adding some more from scratch.
109 */
110 static const uint16_t quantization_per_subband[2][3][13][9] = {
111 {{
112 { 16, 16, 8, 4, 4, 2, 6, 6, 9, }, // film3+
113 { 16, 16, 8, 4, 4, 2, 6, 6, 9, }, // film3
114 { 16, 16, 8, 4, 4, 2, 7, 7, 10, }, // film2+
115 { 16, 16, 8, 4, 4, 2, 8, 8, 12, }, // film2
116 { 16, 16, 8, 4, 4, 2, 16, 16, 26, }, // film1++
117 { 24, 24, 12, 6, 6, 3, 24, 24, 36, }, // film1+
118 { 24, 24, 12, 6, 6, 3, 24, 24, 36, }, // film1
119 { 32, 32, 24, 8, 8, 6, 32, 32, 48, }, // high+
120 { 32, 32, 24, 8, 8, 6, 32, 32, 48, }, // high
121 { 48, 48, 32, 12, 12, 8, 64, 64, 96, }, // medium+
122 { 48, 48, 32, 12, 12, 8, 64, 64, 96, }, // medium
123 { 64, 64, 48, 16, 16, 12, 96, 96, 144, }, // low+
124 { 64, 64, 48, 16, 16, 12, 128, 128, 192, }, // low
125 },
126 {
127 { 16, 16, 8, 4, 4, 2, 6, 6, 9, }, // film3+
128 { 16, 16, 8, 4, 4, 2, 6, 6, 12, }, // film3
129 { 16, 16, 8, 4, 4, 2, 7, 7, 14, }, // film2+
130 { 16, 16, 8, 4, 4, 2, 8, 8, 16, }, // film2
131 { 16, 16, 8, 4, 4, 2, 16, 16, 26, }, // film1++
132 { 24, 24, 12, 6, 6, 3, 24, 24, 36, }, // film1+
133 { 24, 24, 12, 6, 6, 3, 24, 24, 48, }, // film1
134 { 32, 32, 24, 8, 8, 6, 32, 32, 48, }, // high+
135 { 48, 48, 32, 12, 12, 8, 32, 32, 64, }, // high
136 { 48, 48, 32, 12, 12, 8, 64, 64, 96, }, // medium+
137 { 48, 48, 32, 12, 12, 8, 64, 64, 128, }, // medium
138 { 64, 64, 48, 16, 16, 12, 96, 96, 160, }, // low+
139 { 64, 64, 48, 16, 16, 12, 128, 128, 192, }, // low
140 },
141 {
142 { 16, 16, 8, 4, 4, 2, 6, 6, 9, }, // film3+
143 { 16, 16, 8, 4, 4, 2, 6, 6, 12, }, // film3
144 { 16, 16, 8, 4, 4, 2, 7, 7, 14, }, // film2+
145 { 16, 16, 8, 4, 4, 2, 8, 8, 16, }, // film2
146 { 16, 16, 8, 4, 4, 2, 16, 16, 26, }, // film1++
147 { 24, 24, 12, 6, 6, 3, 24, 24, 36, }, // film1+
148 { 24, 24, 12, 6, 6, 3, 24, 24, 48, }, // film1
149 { 32, 32, 24, 8, 8, 6, 32, 32, 48, }, // high+
150 { 48, 48, 32, 12, 12, 8, 32, 32, 64, }, // high
151 { 48, 48, 32, 12, 12, 8, 64, 64, 96, }, // medium+
152 { 48, 48, 32, 12, 12, 8, 64, 64, 128, }, // medium
153 { 64, 64, 48, 16, 16, 12, 96, 96, 160, }, // low+
154 { 64, 64, 48, 16, 16, 12, 128, 128, 192, }, // low
155 }},
156 {{
157 { 16, 16, 8, 16, 16, 8, 24, 24, 36, }, // film3+
158 { 16, 16, 8, 16, 16, 8, 24, 24, 36, }, // film3
159 { 16, 16, 8, 16, 16, 8, 32, 32, 48, }, // film2+
160 { 16, 16, 8, 16, 16, 8, 32, 32, 48, }, // film2
161 { 16, 16, 8, 20, 20, 10, 80, 80, 128, }, // film1++
162 { 24, 24, 12, 24, 24, 12, 96, 96, 144, }, // film1+
163 { 24, 24, 12, 24, 24, 12, 96, 96, 144, }, // film1
164 { 32, 32, 24, 32, 32, 24, 128, 128, 192, }, // high+
165 { 32, 32, 24, 32, 32, 24, 128, 128, 192, }, // high
166 { 48, 48, 32, 48, 48, 32, 256, 256, 384, }, // medium+
167 { 48, 48, 32, 48, 48, 32, 256, 256, 384, }, // medium
168 { 56, 56, 40, 56, 56, 40, 512, 512, 768, }, // low+
169 { 64, 64, 48, 64, 64, 48, 512, 512, 768, }, // low
170 },
171 {
172 { 16, 16, 8, 16, 16, 8, 24, 24, 36, }, // film3+
173 { 16, 16, 8, 16, 16, 8, 48, 48, 72, }, // film3
174 { 16, 16, 8, 16, 16, 8, 48, 48, 72, }, // film2+
175 { 16, 16, 8, 16, 16, 8, 64, 64, 96, }, // film2
176 { 16, 16, 8, 20, 20, 10, 80, 80, 128, }, // film1++
177 { 24, 24, 12, 24, 24, 12, 96, 96, 144, }, // film1+
178 { 24, 24, 12, 24, 24, 12, 192, 192, 288, }, // film1
179 { 32, 32, 24, 32, 32, 24, 128, 128, 192, }, // high+
180 { 32, 32, 24, 32, 32, 24, 256, 256, 384, }, // high
181 { 48, 48, 32, 48, 48, 32, 256, 256, 384, }, // medium+
182 { 48, 48, 32, 48, 48, 32, 512, 512, 768, }, // medium
183 { 56, 56, 40, 56, 56, 40, 512, 512, 768, }, // low+
184 { 64, 64, 48, 64, 64, 48,1024,1024,1536, }, // low
185 },
186 {
187 { 16, 16, 8, 16, 16, 8, 24, 24, 36, }, // film3+
188 { 16, 16, 8, 16, 16, 8, 48, 48, 72, }, // film3
189 { 16, 16, 8, 16, 16, 8, 48, 48, 72, }, // film2+
190 { 16, 16, 8, 16, 16, 8, 64, 64, 96, }, // film2
191 { 16, 16, 10, 20, 20, 10, 80, 80, 128, }, // film1++
192 { 24, 24, 12, 24, 24, 12, 96, 96, 144, }, // film1+
193 { 24, 24, 12, 24, 24, 12, 192, 192, 288, }, // film1
194 { 32, 32, 24, 32, 32, 24, 128, 128, 192, }, // high+
195 { 32, 32, 24, 32, 32, 24, 256, 256, 384, }, // high
196 { 48, 48, 32, 48, 48, 32, 256, 256, 384, }, // medium+
197 { 48, 48, 32, 48, 48, 32, 512, 512, 768, }, // medium
198 { 56, 56, 40, 56, 56, 40, 512, 512, 768, }, // low+
199 { 64, 64, 48, 64, 64, 48,1024,1024,1536, }, // low
200 }},
201 };
202
203 typedef struct Codebook {
204 unsigned bits;
205 unsigned size;
206 } Codebook;
207
208 typedef struct Runbook {
209 unsigned size;
210 unsigned bits;
211 unsigned run;
212 } Runbook;
213
214 typedef struct PlaneEnc {
215 unsigned size;
216
217 int16_t *dwt_buf;
218 int16_t *dwt_tmp;
219
220 unsigned quantization[SUBBAND_COUNT];
221 int16_t *subband[SUBBAND_COUNT];
222 int16_t *l_h[8];
223
224 SubBand band[DWT_LEVELS][4];
225 } PlaneEnc;
226
227 typedef struct CFHDEncContext {
228 const AVClass *class;
229
230 PutBitContext pb;
231 PutByteContext pby;
232
233 int quality;
234 int planes;
235 int chroma_h_shift;
236 int chroma_v_shift;
237 PlaneEnc plane[4];
238
239 uint16_t lut[1024];
240 Runbook rb[321];
241 Codebook cb[513];
242 int16_t *alpha;
243
244 CFHDEncDSPContext dsp;
245 } CFHDEncContext;
246
cfhd_encode_init(AVCodecContext * avctx)247 static av_cold int cfhd_encode_init(AVCodecContext *avctx)
248 {
249 CFHDEncContext *s = avctx->priv_data;
250 const int sign_mask = 256;
251 const int twos_complement = -sign_mask;
252 const int mag_mask = sign_mask - 1;
253 int ret, last = 0;
254
255 ret = av_pix_fmt_get_chroma_sub_sample(avctx->pix_fmt,
256 &s->chroma_h_shift,
257 &s->chroma_v_shift);
258 if (ret < 0)
259 return ret;
260
261 if (avctx->width & 15) {
262 av_log(avctx, AV_LOG_ERROR, "Width must be multiple of 16.\n");
263 return AVERROR_INVALIDDATA;
264 }
265
266 s->planes = av_pix_fmt_count_planes(avctx->pix_fmt);
267
268 for (int i = 0; i < s->planes; i++) {
269 int w8, h8, w4, h4, w2, h2;
270 int width = i ? avctx->width >> s->chroma_h_shift : avctx->width;
271 int height = i ? FFALIGN(avctx->height >> s->chroma_v_shift, 8) :
272 FFALIGN(avctx->height >> s->chroma_v_shift, 8);
273 ptrdiff_t stride = (FFALIGN(width / 8, 8) + 64) * 8;
274
275 w8 = FFALIGN(width / 8, 8) + 64;
276 h8 = FFALIGN(height, 8) / 8;
277 w4 = w8 * 2;
278 h4 = h8 * 2;
279 w2 = w4 * 2;
280 h2 = h4 * 2;
281
282 s->plane[i].dwt_buf =
283 av_mallocz_array(height * stride, sizeof(*s->plane[i].dwt_buf));
284 s->plane[i].dwt_tmp =
285 av_malloc_array(height * stride, sizeof(*s->plane[i].dwt_tmp));
286 if (!s->plane[i].dwt_buf || !s->plane[i].dwt_tmp)
287 return AVERROR(ENOMEM);
288
289 s->plane[i].subband[0] = s->plane[i].dwt_buf;
290 s->plane[i].subband[1] = s->plane[i].dwt_buf + 2 * w8 * h8;
291 s->plane[i].subband[2] = s->plane[i].dwt_buf + 1 * w8 * h8;
292 s->plane[i].subband[3] = s->plane[i].dwt_buf + 3 * w8 * h8;
293 s->plane[i].subband[4] = s->plane[i].dwt_buf + 2 * w4 * h4;
294 s->plane[i].subband[5] = s->plane[i].dwt_buf + 1 * w4 * h4;
295 s->plane[i].subband[6] = s->plane[i].dwt_buf + 3 * w4 * h4;
296 s->plane[i].subband[7] = s->plane[i].dwt_buf + 2 * w2 * h2;
297 s->plane[i].subband[8] = s->plane[i].dwt_buf + 1 * w2 * h2;
298 s->plane[i].subband[9] = s->plane[i].dwt_buf + 3 * w2 * h2;
299
300 for (int j = 0; j < DWT_LEVELS; j++) {
301 for (int k = 0; k < FF_ARRAY_ELEMS(s->plane[i].band[j]); k++) {
302 s->plane[i].band[j][k].width = (width / 8) << j;
303 s->plane[i].band[j][k].height = (height / 8) << j;
304 s->plane[i].band[j][k].a_width = w8 << j;
305 s->plane[i].band[j][k].a_height = h8 << j;
306 }
307 }
308
309 /* ll2 and ll1 commented out because they are done in-place */
310 s->plane[i].l_h[0] = s->plane[i].dwt_tmp;
311 s->plane[i].l_h[1] = s->plane[i].dwt_tmp + 2 * w8 * h8;
312 // s->plane[i].l_h[2] = ll2;
313 s->plane[i].l_h[3] = s->plane[i].dwt_tmp;
314 s->plane[i].l_h[4] = s->plane[i].dwt_tmp + 2 * w4 * h4;
315 // s->plane[i].l_h[5] = ll1;
316 s->plane[i].l_h[6] = s->plane[i].dwt_tmp;
317 s->plane[i].l_h[7] = s->plane[i].dwt_tmp + 2 * w2 * h2;
318 }
319
320 for (int i = 0; i < 512; i++) {
321 int value = (i & sign_mask) ? twos_complement + (i & mag_mask): i;
322 int mag = FFMIN(FFABS(value), 255);
323
324 if (mag) {
325 s->cb[i].bits = (codebook[mag][1] << 1) | (value > 0 ? 0 : 1);
326 s->cb[i].size = codebook[mag][0] + 1;
327 } else {
328 s->cb[i].bits = codebook[mag][1];
329 s->cb[i].size = codebook[mag][0];
330 }
331 }
332
333 s->cb[512].bits = 0x3114ba3;
334 s->cb[512].size = 26;
335
336 s->rb[0].run = 0;
337
338 for (int i = 1, j = 0; i < 320 && j < 17; j++) {
339 int run = runbook[j][2];
340 int end = runbook[j+1][2];
341
342 while (i < end) {
343 s->rb[i].run = run;
344 s->rb[i].bits = runbook[j][1];
345 s->rb[i++].size = runbook[j][0];
346 }
347 }
348
349 s->rb[320].bits = runbook[17][1];
350 s->rb[320].size = runbook[17][0];
351 s->rb[320].run = 320;
352
353 for (int i = 0; i < 256; i++) {
354 int idx = i + ((768LL * i * i * i) / (256 * 256 * 256));
355
356 s->lut[idx] = i;
357 }
358 for (int i = 0; i < 1024; i++) {
359 if (s->lut[i])
360 last = s->lut[i];
361 else
362 s->lut[i] = last;
363 }
364
365 ff_cfhdencdsp_init(&s->dsp);
366
367 if (s->planes != 4)
368 return 0;
369
370 s->alpha = av_calloc(avctx->width * avctx->height, sizeof(*s->alpha));
371 if (!s->alpha)
372 return AVERROR(ENOMEM);
373
374 return 0;
375 }
376
quantize_band(int16_t * input,int width,int a_width,int height,unsigned quantization)377 static void quantize_band(int16_t *input, int width, int a_width,
378 int height, unsigned quantization)
379 {
380 const int16_t factor = (uint32_t)(1U << 15) / quantization;
381
382 for (int i = 0; i < height; i++) {
383 for (int j = 0; j < width; j++)
384 input[j] = av_clip_intp2(((input[j] * factor + 16384 * FFSIGN(input[j])) / 32768), 10);
385 input += a_width;
386 }
387 }
388
put_runcode(PutBitContext * pb,int count,const Runbook * const rb)389 static int put_runcode(PutBitContext *pb, int count, const Runbook *const rb)
390 {
391 while (count > 0) {
392 const int index = FFMIN(320, count);
393
394 put_bits(pb, rb[index].size, rb[index].bits);
395 count -= rb[index].run;
396 }
397
398 return 0;
399 }
400
process_alpha(const int16_t * src,int width,int height,ptrdiff_t stride,int16_t * dst)401 static void process_alpha(const int16_t *src, int width, int height, ptrdiff_t stride, int16_t *dst)
402 {
403 for (int i = 0; i < height; i++) {
404 for (int j = 0; j < width; j++) {
405 int alpha = src[j];
406
407 if (alpha > 0 && alpha < 4080) {
408 alpha *= 223;
409 alpha += 128;
410 alpha >>= 8;
411 alpha += 256;
412 }
413
414 dst[j] = av_clip_uintp2(alpha, 12);
415 }
416
417 src += stride;
418 dst += width;
419 }
420 }
421
cfhd_encode_frame(AVCodecContext * avctx,AVPacket * pkt,const AVFrame * frame,int * got_packet)422 static int cfhd_encode_frame(AVCodecContext *avctx, AVPacket *pkt,
423 const AVFrame *frame, int *got_packet)
424 {
425 CFHDEncContext *s = avctx->priv_data;
426 CFHDEncDSPContext *dsp = &s->dsp;
427 PutByteContext *pby = &s->pby;
428 PutBitContext *pb = &s->pb;
429 const Codebook *const cb = s->cb;
430 const Runbook *const rb = s->rb;
431 const uint16_t *lut = s->lut;
432 unsigned pos;
433 int ret;
434
435 for (int plane = 0; plane < s->planes; plane++) {
436 int width = s->plane[plane].band[2][0].width;
437 int a_width = s->plane[plane].band[2][0].a_width;
438 int height = s->plane[plane].band[2][0].height;
439 int act_plane = plane == 1 ? 2 : plane == 2 ? 1 : plane;
440 int16_t *input = (int16_t *)frame->data[act_plane];
441 int16_t *low = s->plane[plane].l_h[6];
442 int16_t *high = s->plane[plane].l_h[7];
443 ptrdiff_t in_stride = frame->linesize[act_plane] / 2;
444 int low_stride, high_stride;
445
446 if (plane == 3) {
447 process_alpha(input, avctx->width, avctx->height,
448 in_stride, s->alpha);
449 input = s->alpha;
450 in_stride = avctx->width;
451 }
452
453 dsp->horiz_filter(input, low, high,
454 in_stride, a_width, a_width,
455 width * 2, height * 2);
456
457 input = s->plane[plane].l_h[7];
458 low = s->plane[plane].subband[7];
459 low_stride = s->plane[plane].band[2][0].a_width;
460 high = s->plane[plane].subband[9];
461 high_stride = s->plane[plane].band[2][0].a_width;
462
463 dsp->vert_filter(input, low, high,
464 a_width, low_stride, high_stride,
465 width, height * 2);
466
467 input = s->plane[plane].l_h[6];
468 low = s->plane[plane].l_h[7];
469 high = s->plane[plane].subband[8];
470
471 dsp->vert_filter(input, low, high,
472 a_width, low_stride, high_stride,
473 width, height * 2);
474
475 a_width = s->plane[plane].band[1][0].a_width;
476 width = s->plane[plane].band[1][0].width;
477 height = s->plane[plane].band[1][0].height;
478 input = s->plane[plane].l_h[7];
479 low = s->plane[plane].l_h[3];
480 low_stride = s->plane[plane].band[1][0].a_width;
481 high = s->plane[plane].l_h[4];
482 high_stride = s->plane[plane].band[1][0].a_width;
483
484 for (int i = 0; i < height * 2; i++) {
485 for (int j = 0; j < width * 2; j++)
486 input[j] /= 4;
487 input += a_width * 2;
488 }
489
490 input = s->plane[plane].l_h[7];
491 dsp->horiz_filter(input, low, high,
492 a_width * 2, low_stride, high_stride,
493 width * 2, height * 2);
494
495 input = s->plane[plane].l_h[4];
496 low = s->plane[plane].subband[4];
497 high = s->plane[plane].subband[6];
498
499 dsp->vert_filter(input, low, high,
500 a_width, low_stride, high_stride,
501 width, height * 2);
502
503 input = s->plane[plane].l_h[3];
504 low = s->plane[plane].l_h[4];
505 high = s->plane[plane].subband[5];
506
507 dsp->vert_filter(input, low, high,
508 a_width, low_stride, high_stride,
509 width, height * 2);
510
511 a_width = s->plane[plane].band[0][0].a_width;
512 width = s->plane[plane].band[0][0].width;
513 height = s->plane[plane].band[0][0].height;
514 input = s->plane[plane].l_h[4];
515 low = s->plane[plane].l_h[0];
516 low_stride = s->plane[plane].band[0][0].a_width;
517 high = s->plane[plane].l_h[1];
518 high_stride = s->plane[plane].band[0][0].a_width;
519
520 if (avctx->pix_fmt != AV_PIX_FMT_YUV422P10) {
521 for (int i = 0; i < height * 2; i++) {
522 for (int j = 0; j < width * 2; j++)
523 input[j] /= 4;
524 input += a_width * 2;
525 }
526 }
527
528 input = s->plane[plane].l_h[4];
529 dsp->horiz_filter(input, low, high,
530 a_width * 2, low_stride, high_stride,
531 width * 2, height * 2);
532
533 low = s->plane[plane].subband[1];
534 high = s->plane[plane].subband[3];
535 input = s->plane[plane].l_h[1];
536
537 dsp->vert_filter(input, low, high,
538 a_width, low_stride, high_stride,
539 width, height * 2);
540
541 low = s->plane[plane].subband[0];
542 high = s->plane[plane].subband[2];
543 input = s->plane[plane].l_h[0];
544
545 dsp->vert_filter(input, low, high,
546 a_width, low_stride, high_stride,
547 width, height * 2);
548 }
549
550 ret = ff_alloc_packet2(avctx, pkt, 64LL + s->planes * (2LL * avctx->width * avctx->height + 1000LL), 0);
551 if (ret < 0)
552 return ret;
553
554 bytestream2_init_writer(pby, pkt->data, pkt->size);
555
556 bytestream2_put_be16(pby, SampleType);
557 bytestream2_put_be16(pby, 9);
558
559 bytestream2_put_be16(pby, SampleIndexTable);
560 bytestream2_put_be16(pby, s->planes);
561
562 for (int i = 0; i < s->planes; i++)
563 bytestream2_put_be32(pby, 0);
564
565 bytestream2_put_be16(pby, TransformType);
566 bytestream2_put_be16(pby, 0);
567
568 bytestream2_put_be16(pby, NumFrames);
569 bytestream2_put_be16(pby, 1);
570
571 bytestream2_put_be16(pby, ChannelCount);
572 bytestream2_put_be16(pby, s->planes);
573
574 bytestream2_put_be16(pby, EncodedFormat);
575 bytestream2_put_be16(pby, avctx->pix_fmt == AV_PIX_FMT_YUV422P10 ? 1 : 3 + (s->planes == 4));
576
577 bytestream2_put_be16(pby, WaveletCount);
578 bytestream2_put_be16(pby, 3);
579
580 bytestream2_put_be16(pby, SubbandCount);
581 bytestream2_put_be16(pby, SUBBAND_COUNT);
582
583 bytestream2_put_be16(pby, NumSpatial);
584 bytestream2_put_be16(pby, 2);
585
586 bytestream2_put_be16(pby, FirstWavelet);
587 bytestream2_put_be16(pby, 3);
588
589 bytestream2_put_be16(pby, ImageWidth);
590 bytestream2_put_be16(pby, avctx->width);
591
592 bytestream2_put_be16(pby, ImageHeight);
593 bytestream2_put_be16(pby, avctx->height);
594
595 bytestream2_put_be16(pby, -FrameNumber);
596 bytestream2_put_be16(pby, frame->pts & 0xFFFF);
597
598 bytestream2_put_be16(pby, Precision);
599 bytestream2_put_be16(pby, avctx->pix_fmt == AV_PIX_FMT_YUV422P10 ? 10 : 12);
600
601 bytestream2_put_be16(pby, PrescaleTable);
602 bytestream2_put_be16(pby, avctx->pix_fmt == AV_PIX_FMT_YUV422P10 ? 0x2000 : 0x2800);
603
604 bytestream2_put_be16(pby, SampleFlags);
605 bytestream2_put_be16(pby, 1);
606
607 for (int p = 0; p < s->planes; p++) {
608 int width = s->plane[p].band[0][0].width;
609 int a_width = s->plane[p].band[0][0].a_width;
610 int height = s->plane[p].band[0][0].height;
611 int16_t *data = s->plane[p].subband[0];
612
613 if (p) {
614 bytestream2_put_be16(pby, SampleType);
615 bytestream2_put_be16(pby, 3);
616
617 bytestream2_put_be16(pby, ChannelNumber);
618 bytestream2_put_be16(pby, p);
619 }
620
621 bytestream2_put_be16(pby, BitstreamMarker);
622 bytestream2_put_be16(pby, 0x1a4a);
623
624 pos = bytestream2_tell_p(pby);
625
626 bytestream2_put_be16(pby, LowpassSubband);
627 bytestream2_put_be16(pby, 0);
628
629 bytestream2_put_be16(pby, NumLevels);
630 bytestream2_put_be16(pby, 3);
631
632 bytestream2_put_be16(pby, LowpassWidth);
633 bytestream2_put_be16(pby, width);
634
635 bytestream2_put_be16(pby, LowpassHeight);
636 bytestream2_put_be16(pby, height);
637
638 bytestream2_put_be16(pby, PixelOffset);
639 bytestream2_put_be16(pby, 0);
640
641 bytestream2_put_be16(pby, LowpassQuantization);
642 bytestream2_put_be16(pby, 1);
643
644 bytestream2_put_be16(pby, LowpassPrecision);
645 bytestream2_put_be16(pby, 16);
646
647 bytestream2_put_be16(pby, BitstreamMarker);
648 bytestream2_put_be16(pby, 0x0f0f);
649
650 for (int i = 0; i < height; i++) {
651 for (int j = 0; j < width; j++)
652 bytestream2_put_be16(pby, data[j]);
653 data += a_width;
654 }
655
656 bytestream2_put_be16(pby, BitstreamMarker);
657 bytestream2_put_be16(pby, 0x1b4b);
658
659 for (int l = 0; l < 3; l++) {
660 for (int i = 0; i < 3; i++) {
661 s->plane[p].quantization[1 + l * 3 + i] = quantization_per_subband[avctx->pix_fmt != AV_PIX_FMT_YUV422P10][p >= 3 ? 0 : p][s->quality][l * 3 + i];
662 }
663 }
664
665 for (int l = 0; l < 3; l++) {
666 int a_width = s->plane[p].band[l][0].a_width;
667 int width = s->plane[p].band[l][0].width;
668 int stride = FFALIGN(width, 8);
669 int height = s->plane[p].band[l][0].height;
670
671 bytestream2_put_be16(pby, BitstreamMarker);
672 bytestream2_put_be16(pby, 0x0d0d);
673
674 bytestream2_put_be16(pby, WaveletType);
675 bytestream2_put_be16(pby, 3 + 2 * (l == 2));
676
677 bytestream2_put_be16(pby, WaveletNumber);
678 bytestream2_put_be16(pby, 3 - l);
679
680 bytestream2_put_be16(pby, WaveletLevel);
681 bytestream2_put_be16(pby, 3 - l);
682
683 bytestream2_put_be16(pby, NumBands);
684 bytestream2_put_be16(pby, 4);
685
686 bytestream2_put_be16(pby, HighpassWidth);
687 bytestream2_put_be16(pby, width);
688
689 bytestream2_put_be16(pby, HighpassHeight);
690 bytestream2_put_be16(pby, height);
691
692 bytestream2_put_be16(pby, LowpassBorder);
693 bytestream2_put_be16(pby, 0);
694
695 bytestream2_put_be16(pby, HighpassBorder);
696 bytestream2_put_be16(pby, 0);
697
698 bytestream2_put_be16(pby, LowpassScale);
699 bytestream2_put_be16(pby, 1);
700
701 bytestream2_put_be16(pby, LowpassDivisor);
702 bytestream2_put_be16(pby, 1);
703
704 for (int i = 0; i < 3; i++) {
705 int16_t *data = s->plane[p].subband[1 + l * 3 + i];
706 int count = 0, padd = 0;
707
708 bytestream2_put_be16(pby, BitstreamMarker);
709 bytestream2_put_be16(pby, 0x0e0e);
710
711 bytestream2_put_be16(pby, SubbandNumber);
712 bytestream2_put_be16(pby, i + 1);
713
714 bytestream2_put_be16(pby, BandCodingFlags);
715 bytestream2_put_be16(pby, 1);
716
717 bytestream2_put_be16(pby, BandWidth);
718 bytestream2_put_be16(pby, width);
719
720 bytestream2_put_be16(pby, BandHeight);
721 bytestream2_put_be16(pby, height);
722
723 bytestream2_put_be16(pby, SubbandBand);
724 bytestream2_put_be16(pby, 1 + l * 3 + i);
725
726 bytestream2_put_be16(pby, BandEncoding);
727 bytestream2_put_be16(pby, 3);
728
729 bytestream2_put_be16(pby, Quantization);
730 bytestream2_put_be16(pby, s->plane[p].quantization[1 + l * 3 + i]);
731
732 bytestream2_put_be16(pby, BandScale);
733 bytestream2_put_be16(pby, 1);
734
735 bytestream2_put_be16(pby, BandHeader);
736 bytestream2_put_be16(pby, 0);
737
738 quantize_band(data, width, a_width, height,
739 s->plane[p].quantization[1 + l * 3 + i]);
740
741 init_put_bits(pb, pkt->data + bytestream2_tell_p(pby), bytestream2_get_bytes_left_p(pby));
742
743 for (int m = 0; m < height; m++) {
744 for (int j = 0; j < stride; j++) {
745 int16_t index = j >= width ? 0 : FFSIGN(data[j]) * lut[FFABS(data[j])];
746
747 if (index < 0)
748 index += 512;
749 if (index == 0) {
750 count++;
751 continue;
752 } else if (count > 0) {
753 count = put_runcode(pb, count, rb);
754 }
755
756 put_bits(pb, cb[index].size, cb[index].bits);
757 }
758
759 data += a_width;
760 }
761
762 if (count > 0) {
763 count = put_runcode(pb, count, rb);
764 }
765
766 put_bits(pb, cb[512].size, cb[512].bits);
767
768 flush_put_bits(pb);
769 bytestream2_skip_p(pby, put_bits_count(pb) >> 3);
770 padd = (4 - (bytestream2_tell_p(pby) & 3)) & 3;
771 while (padd--)
772 bytestream2_put_byte(pby, 0);
773
774 bytestream2_put_be16(pby, BandTrailer);
775 bytestream2_put_be16(pby, 0);
776 }
777
778 bytestream2_put_be16(pby, BitstreamMarker);
779 bytestream2_put_be16(pby, 0x0c0c);
780 }
781
782 s->plane[p].size = bytestream2_tell_p(pby) - pos;
783 }
784
785 bytestream2_put_be16(pby, GroupTrailer);
786 bytestream2_put_be16(pby, 0);
787
788 av_shrink_packet(pkt, bytestream2_tell_p(pby));
789
790 pkt->flags |= AV_PKT_FLAG_KEY;
791
792 bytestream2_seek_p(pby, 8, SEEK_SET);
793 for (int i = 0; i < s->planes; i++)
794 bytestream2_put_be32(pby, s->plane[i].size);
795
796 *got_packet = 1;
797
798 return 0;
799 }
800
cfhd_encode_close(AVCodecContext * avctx)801 static av_cold int cfhd_encode_close(AVCodecContext *avctx)
802 {
803 CFHDEncContext *s = avctx->priv_data;
804
805 for (int i = 0; i < s->planes; i++) {
806 av_freep(&s->plane[i].dwt_buf);
807 av_freep(&s->plane[i].dwt_tmp);
808
809 for (int j = 0; j < SUBBAND_COUNT; j++)
810 s->plane[i].subband[j] = NULL;
811
812 for (int j = 0; j < 8; j++)
813 s->plane[i].l_h[j] = NULL;
814 }
815
816 av_freep(&s->alpha);
817
818 return 0;
819 }
820
821 #define OFFSET(x) offsetof(CFHDEncContext, x)
822 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
823 static const AVOption options[] = {
824 { "quality", "set quality", OFFSET(quality), AV_OPT_TYPE_INT, {.i64= 0}, 0, 12, VE, "q" },
825 { "film3+", NULL, 0, AV_OPT_TYPE_CONST, {.i64= 0}, 0, 0, VE, "q" },
826 { "film3", NULL, 0, AV_OPT_TYPE_CONST, {.i64= 1}, 0, 0, VE, "q" },
827 { "film2+", NULL, 0, AV_OPT_TYPE_CONST, {.i64= 2}, 0, 0, VE, "q" },
828 { "film2", NULL, 0, AV_OPT_TYPE_CONST, {.i64= 3}, 0, 0, VE, "q" },
829 { "film1.5", NULL, 0, AV_OPT_TYPE_CONST, {.i64= 4}, 0, 0, VE, "q" },
830 { "film1+", NULL, 0, AV_OPT_TYPE_CONST, {.i64= 5}, 0, 0, VE, "q" },
831 { "film1", NULL, 0, AV_OPT_TYPE_CONST, {.i64= 6}, 0, 0, VE, "q" },
832 { "high+", NULL, 0, AV_OPT_TYPE_CONST, {.i64= 7}, 0, 0, VE, "q" },
833 { "high", NULL, 0, AV_OPT_TYPE_CONST, {.i64= 8}, 0, 0, VE, "q" },
834 { "medium+", NULL, 0, AV_OPT_TYPE_CONST, {.i64= 9}, 0, 0, VE, "q" },
835 { "medium", NULL, 0, AV_OPT_TYPE_CONST, {.i64=10}, 0, 0, VE, "q" },
836 { "low+", NULL, 0, AV_OPT_TYPE_CONST, {.i64=11}, 0, 0, VE, "q" },
837 { "low", NULL, 0, AV_OPT_TYPE_CONST, {.i64=12}, 0, 0, VE, "q" },
838 { NULL},
839 };
840
841 static const AVClass cfhd_class = {
842 .class_name = "cfhd",
843 .item_name = av_default_item_name,
844 .option = options,
845 .version = LIBAVUTIL_VERSION_INT,
846 };
847
848 AVCodec ff_cfhd_encoder = {
849 .name = "cfhd",
850 .long_name = NULL_IF_CONFIG_SMALL("GoPro CineForm HD"),
851 .type = AVMEDIA_TYPE_VIDEO,
852 .id = AV_CODEC_ID_CFHD,
853 .priv_data_size = sizeof(CFHDEncContext),
854 .priv_class = &cfhd_class,
855 .init = cfhd_encode_init,
856 .close = cfhd_encode_close,
857 .encode2 = cfhd_encode_frame,
858 .capabilities = AV_CODEC_CAP_FRAME_THREADS,
859 .pix_fmts = (const enum AVPixelFormat[]) {
860 AV_PIX_FMT_YUV422P10,
861 AV_PIX_FMT_GBRP12,
862 AV_PIX_FMT_GBRAP12,
863 AV_PIX_FMT_NONE
864 },
865 .caps_internal = FF_CODEC_CAP_INIT_CLEANUP,
866 };
867