1 /*
2 * software YUV to RGB converter
3 *
4 * Copyright (C) 2009 Konstantin Shishkov
5 *
6 * 1,4,8bpp support and context / deglobalize stuff
7 * by Michael Niedermayer (michaelni@gmx.at)
8 *
9 * This file is part of FFmpeg.
10 *
11 * FFmpeg is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU Lesser General Public
13 * License as published by the Free Software Foundation; either
14 * version 2.1 of the License, or (at your option) any later version.
15 *
16 * FFmpeg is distributed in the hope that it will be useful,
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
19 * Lesser General Public License for more details.
20 *
21 * You should have received a copy of the GNU Lesser General Public
22 * License along with FFmpeg; if not, write to the Free Software
23 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 */
25
26 #include <stdio.h>
27 #include <stdlib.h>
28 #include <inttypes.h>
29
30 #include "libavutil/cpu.h"
31 #include "libavutil/bswap.h"
32
33 #include "config.h"
34
35 #include "rgb2rgb.h"
36 #include "swscale.h"
37 #include "swscale_internal.h"
38 #include "libavutil/pixdesc.h"
39
40 const int32_t ff_yuv2rgb_coeffs[8][4] = {
41 { 117504, 138453, 13954, 34903 }, /* no sequence_display_extension */
42 { 117504, 138453, 13954, 34903 }, /* ITU-R Rec. 709 (1990) */
43 { 104597, 132201, 25675, 53279 }, /* unspecified */
44 { 104597, 132201, 25675, 53279 }, /* reserved */
45 { 104448, 132798, 24759, 53109 }, /* FCC */
46 { 104597, 132201, 25675, 53279 }, /* ITU-R Rec. 624-4 System B, G */
47 { 104597, 132201, 25675, 53279 }, /* SMPTE 170M */
48 { 117579, 136230, 16907, 35559 } /* SMPTE 240M (1987) */
49 };
50
sws_getCoefficients(int colorspace)51 const int *sws_getCoefficients(int colorspace)
52 {
53 if (colorspace > 7 || colorspace < 0)
54 colorspace = SWS_CS_DEFAULT;
55 return ff_yuv2rgb_coeffs[colorspace];
56 }
57
58 #define LOADCHROMA(i) \
59 U = pu[i]; \
60 V = pv[i]; \
61 r = (void *)c->table_rV[V+YUVRGB_TABLE_HEADROOM]; \
62 g = (void *)(c->table_gU[U+YUVRGB_TABLE_HEADROOM] + c->table_gV[V+YUVRGB_TABLE_HEADROOM]); \
63 b = (void *)c->table_bU[U+YUVRGB_TABLE_HEADROOM];
64
65 #define PUTRGB(dst, src, i) \
66 Y = src[2 * i]; \
67 dst[2 * i] = r[Y] + g[Y] + b[Y]; \
68 Y = src[2 * i + 1]; \
69 dst[2 * i + 1] = r[Y] + g[Y] + b[Y];
70
71 #define PUTRGB24(dst, src, i) \
72 Y = src[2 * i]; \
73 dst[6 * i + 0] = r[Y]; \
74 dst[6 * i + 1] = g[Y]; \
75 dst[6 * i + 2] = b[Y]; \
76 Y = src[2 * i + 1]; \
77 dst[6 * i + 3] = r[Y]; \
78 dst[6 * i + 4] = g[Y]; \
79 dst[6 * i + 5] = b[Y];
80
81 #define PUTBGR24(dst, src, i) \
82 Y = src[2 * i]; \
83 dst[6 * i + 0] = b[Y]; \
84 dst[6 * i + 1] = g[Y]; \
85 dst[6 * i + 2] = r[Y]; \
86 Y = src[2 * i + 1]; \
87 dst[6 * i + 3] = b[Y]; \
88 dst[6 * i + 4] = g[Y]; \
89 dst[6 * i + 5] = r[Y];
90
91 #define PUTRGBA(dst, ysrc, asrc, i, s) \
92 Y = ysrc[2 * i]; \
93 dst[2 * i] = r[Y] + g[Y] + b[Y] + (asrc[2 * i] << s); \
94 Y = ysrc[2 * i + 1]; \
95 dst[2 * i + 1] = r[Y] + g[Y] + b[Y] + (asrc[2 * i + 1] << s);
96
97 #define PUTRGB48(dst, src, i) \
98 Y = src[ 2 * i]; \
99 dst[12 * i + 0] = dst[12 * i + 1] = r[Y]; \
100 dst[12 * i + 2] = dst[12 * i + 3] = g[Y]; \
101 dst[12 * i + 4] = dst[12 * i + 5] = b[Y]; \
102 Y = src[ 2 * i + 1]; \
103 dst[12 * i + 6] = dst[12 * i + 7] = r[Y]; \
104 dst[12 * i + 8] = dst[12 * i + 9] = g[Y]; \
105 dst[12 * i + 10] = dst[12 * i + 11] = b[Y];
106
107 #define PUTBGR48(dst, src, i) \
108 Y = src[2 * i]; \
109 dst[12 * i + 0] = dst[12 * i + 1] = b[Y]; \
110 dst[12 * i + 2] = dst[12 * i + 3] = g[Y]; \
111 dst[12 * i + 4] = dst[12 * i + 5] = r[Y]; \
112 Y = src[2 * i + 1]; \
113 dst[12 * i + 6] = dst[12 * i + 7] = b[Y]; \
114 dst[12 * i + 8] = dst[12 * i + 9] = g[Y]; \
115 dst[12 * i + 10] = dst[12 * i + 11] = r[Y];
116
117 #define YUV2RGBFUNC(func_name, dst_type, alpha) \
118 static int func_name(SwsContext *c, const uint8_t *src[], \
119 int srcStride[], int srcSliceY, int srcSliceH, \
120 uint8_t *dst[], int dstStride[]) \
121 { \
122 int y; \
123 \
124 if (!alpha && c->srcFormat == AV_PIX_FMT_YUV422P) { \
125 srcStride[1] *= 2; \
126 srcStride[2] *= 2; \
127 } \
128 for (y = 0; y < srcSliceH; y += 2) { \
129 dst_type *dst_1 = \
130 (dst_type *)(dst[0] + (y + srcSliceY) * dstStride[0]); \
131 dst_type *dst_2 = \
132 (dst_type *)(dst[0] + (y + srcSliceY + 1) * dstStride[0]); \
133 dst_type av_unused *r, *g, *b; \
134 const uint8_t *py_1 = src[0] + y * srcStride[0]; \
135 const uint8_t *py_2 = py_1 + srcStride[0]; \
136 const uint8_t *pu = src[1] + (y >> 1) * srcStride[1]; \
137 const uint8_t *pv = src[2] + (y >> 1) * srcStride[2]; \
138 const uint8_t av_unused *pa_1, *pa_2; \
139 unsigned int h_size = c->dstW >> 3; \
140 if (alpha) { \
141 pa_1 = src[3] + y * srcStride[3]; \
142 pa_2 = pa_1 + srcStride[3]; \
143 } \
144 while (h_size--) { \
145 int av_unused U, V, Y; \
146
147 #define ENDYUV2RGBLINE(dst_delta, ss) \
148 pu += 4 >> ss; \
149 pv += 4 >> ss; \
150 py_1 += 8 >> ss; \
151 py_2 += 8 >> ss; \
152 dst_1 += dst_delta >> ss; \
153 dst_2 += dst_delta >> ss; \
154 } \
155 if (c->dstW & (4 >> ss)) { \
156 int av_unused Y, U, V; \
157
158 #define ENDYUV2RGBFUNC() \
159 } \
160 } \
161 return srcSliceH; \
162 }
163
164 #define CLOSEYUV2RGBFUNC(dst_delta) \
165 ENDYUV2RGBLINE(dst_delta, 0) \
166 ENDYUV2RGBFUNC()
167
168 YUV2RGBFUNC(yuv2rgb_c_48, uint8_t, 0)
169 LOADCHROMA(0);
170 PUTRGB48(dst_1, py_1, 0);
171 PUTRGB48(dst_2, py_2, 0);
172
173 LOADCHROMA(1);
174 PUTRGB48(dst_2, py_2, 1);
175 PUTRGB48(dst_1, py_1, 1);
176
177 LOADCHROMA(2);
178 PUTRGB48(dst_1, py_1, 2);
179 PUTRGB48(dst_2, py_2, 2);
180
181 LOADCHROMA(3);
182 PUTRGB48(dst_2, py_2, 3);
183 PUTRGB48(dst_1, py_1, 3);
184 ENDYUV2RGBLINE(48, 0)
185 LOADCHROMA(0);
186 PUTRGB48(dst_1, py_1, 0);
187 PUTRGB48(dst_2, py_2, 0);
188
189 LOADCHROMA(1);
190 PUTRGB48(dst_2, py_2, 1);
191 PUTRGB48(dst_1, py_1, 1);
192 ENDYUV2RGBLINE(48, 1)
193 LOADCHROMA(0);
194 PUTRGB48(dst_1, py_1, 0);
195 PUTRGB48(dst_2, py_2, 0);
196 ENDYUV2RGBFUNC()
197
198 YUV2RGBFUNC(yuv2rgb_c_bgr48, uint8_t, 0)
199 LOADCHROMA(0);
200 PUTBGR48(dst_1, py_1, 0);
201 PUTBGR48(dst_2, py_2, 0);
202
203 LOADCHROMA(1);
204 PUTBGR48(dst_2, py_2, 1);
205 PUTBGR48(dst_1, py_1, 1);
206
207 LOADCHROMA(2);
208 PUTBGR48(dst_1, py_1, 2);
209 PUTBGR48(dst_2, py_2, 2);
210
211 LOADCHROMA(3);
212 PUTBGR48(dst_2, py_2, 3);
213 PUTBGR48(dst_1, py_1, 3);
214 ENDYUV2RGBLINE(48, 0)
215 LOADCHROMA(0);
216 PUTBGR48(dst_1, py_1, 0);
217 PUTBGR48(dst_2, py_2, 0);
218
219 LOADCHROMA(1);
220 PUTBGR48(dst_2, py_2, 1);
221 PUTBGR48(dst_1, py_1, 1);
222 ENDYUV2RGBLINE(48, 1)
223 LOADCHROMA(0);
224 PUTBGR48(dst_1, py_1, 0);
225 PUTBGR48(dst_2, py_2, 0);
226 ENDYUV2RGBFUNC()
227
228 YUV2RGBFUNC(yuv2rgb_c_32, uint32_t, 0)
229 LOADCHROMA(0);
230 PUTRGB(dst_1, py_1, 0);
231 PUTRGB(dst_2, py_2, 0);
232
233 LOADCHROMA(1);
234 PUTRGB(dst_2, py_2, 1);
235 PUTRGB(dst_1, py_1, 1);
236
237 LOADCHROMA(2);
238 PUTRGB(dst_1, py_1, 2);
239 PUTRGB(dst_2, py_2, 2);
240
241 LOADCHROMA(3);
242 PUTRGB(dst_2, py_2, 3);
243 PUTRGB(dst_1, py_1, 3);
244 ENDYUV2RGBLINE(8, 0)
245 LOADCHROMA(0);
246 PUTRGB(dst_1, py_1, 0);
247 PUTRGB(dst_2, py_2, 0);
248
249 LOADCHROMA(1);
250 PUTRGB(dst_2, py_2, 1);
251 PUTRGB(dst_1, py_1, 1);
252 ENDYUV2RGBLINE(8, 1)
253 LOADCHROMA(0);
254 PUTRGB(dst_1, py_1, 0);
255 PUTRGB(dst_2, py_2, 0);
256 ENDYUV2RGBFUNC()
257
258 YUV2RGBFUNC(yuva2rgba_c, uint32_t, 1)
259 LOADCHROMA(0);
260 PUTRGBA(dst_1, py_1, pa_1, 0, 24);
261 PUTRGBA(dst_2, py_2, pa_2, 0, 24);
262
263 LOADCHROMA(1);
264 PUTRGBA(dst_2, py_2, pa_2, 1, 24);
265 PUTRGBA(dst_1, py_1, pa_1, 1, 24);
266
267 LOADCHROMA(2);
268 PUTRGBA(dst_1, py_1, pa_1, 2, 24);
269 PUTRGBA(dst_2, py_2, pa_2, 2, 24);
270
271 LOADCHROMA(3);
272 PUTRGBA(dst_2, py_2, pa_2, 3, 24);
273 PUTRGBA(dst_1, py_1, pa_1, 3, 24);
274 pa_1 += 8;
275 pa_2 += 8;
276 ENDYUV2RGBLINE(8, 0)
277 LOADCHROMA(0);
278 PUTRGBA(dst_1, py_1, pa_1, 0, 24);
279 PUTRGBA(dst_2, py_2, pa_2, 0, 24);
280
281 LOADCHROMA(1);
282 PUTRGBA(dst_2, py_2, pa_2, 1, 24);
283 PUTRGBA(dst_1, py_1, pa_1, 1, 24);
284 pa_1 += 4;
285 pa_2 += 4;
286 ENDYUV2RGBLINE(8, 1)
287 LOADCHROMA(0);
288 PUTRGBA(dst_1, py_1, pa_1, 0, 24);
289 PUTRGBA(dst_2, py_2, pa_2, 0, 24);
290 ENDYUV2RGBFUNC()
291
292 YUV2RGBFUNC(yuva2argb_c, uint32_t, 1)
293 LOADCHROMA(0);
294 PUTRGBA(dst_1, py_1, pa_1, 0, 0);
295 PUTRGBA(dst_2, py_2, pa_2, 0, 0);
296
297 LOADCHROMA(1);
298 PUTRGBA(dst_2, py_2, pa_2, 1, 0);
299 PUTRGBA(dst_1, py_1, pa_1, 1, 0);
300
301 LOADCHROMA(2);
302 PUTRGBA(dst_1, py_1, pa_1, 2, 0);
303 PUTRGBA(dst_2, py_2, pa_2, 2, 0);
304
305 LOADCHROMA(3);
306 PUTRGBA(dst_2, py_2, pa_2, 3, 0);
307 PUTRGBA(dst_1, py_1, pa_1, 3, 0);
308 pa_1 += 8;
309 pa_2 += 8;
310 ENDYUV2RGBLINE(8, 0)
311 LOADCHROMA(0);
312 PUTRGBA(dst_1, py_1, pa_1, 0, 0);
313 PUTRGBA(dst_2, py_2, pa_2, 0, 0);
314
315 LOADCHROMA(1);
316 PUTRGBA(dst_2, py_2, pa_2, 1, 0);
317 PUTRGBA(dst_1, py_1, pa_1, 1, 0);
318 pa_1 += 4;
319 pa_2 += 4;
320 ENDYUV2RGBLINE(8, 1)
321 LOADCHROMA(0);
322 PUTRGBA(dst_1, py_1, pa_1, 0, 0);
323 PUTRGBA(dst_2, py_2, pa_2, 0, 0);
324 ENDYUV2RGBFUNC()
325
326 YUV2RGBFUNC(yuv2rgb_c_24_rgb, uint8_t, 0)
327 LOADCHROMA(0);
328 PUTRGB24(dst_1, py_1, 0);
329 PUTRGB24(dst_2, py_2, 0);
330
331 LOADCHROMA(1);
332 PUTRGB24(dst_2, py_2, 1);
333 PUTRGB24(dst_1, py_1, 1);
334
335 LOADCHROMA(2);
336 PUTRGB24(dst_1, py_1, 2);
337 PUTRGB24(dst_2, py_2, 2);
338
339 LOADCHROMA(3);
340 PUTRGB24(dst_2, py_2, 3);
341 PUTRGB24(dst_1, py_1, 3);
342 ENDYUV2RGBLINE(24, 0)
343 LOADCHROMA(0);
344 PUTRGB24(dst_1, py_1, 0);
345 PUTRGB24(dst_2, py_2, 0);
346
347 LOADCHROMA(1);
348 PUTRGB24(dst_2, py_2, 1);
349 PUTRGB24(dst_1, py_1, 1);
350 ENDYUV2RGBLINE(24, 1)
351 LOADCHROMA(0);
352 PUTRGB24(dst_1, py_1, 0);
353 PUTRGB24(dst_2, py_2, 0);
354 ENDYUV2RGBFUNC()
355
356 // only trivial mods from yuv2rgb_c_24_rgb
357 YUV2RGBFUNC(yuv2rgb_c_24_bgr, uint8_t, 0)
358 LOADCHROMA(0);
359 PUTBGR24(dst_1, py_1, 0);
360 PUTBGR24(dst_2, py_2, 0);
361
362 LOADCHROMA(1);
363 PUTBGR24(dst_2, py_2, 1);
364 PUTBGR24(dst_1, py_1, 1);
365
366 LOADCHROMA(2);
367 PUTBGR24(dst_1, py_1, 2);
368 PUTBGR24(dst_2, py_2, 2);
369
370 LOADCHROMA(3);
371 PUTBGR24(dst_2, py_2, 3);
372 PUTBGR24(dst_1, py_1, 3);
373 ENDYUV2RGBLINE(24, 0)
374 LOADCHROMA(0);
375 PUTBGR24(dst_1, py_1, 0);
376 PUTBGR24(dst_2, py_2, 0);
377
378 LOADCHROMA(1);
379 PUTBGR24(dst_2, py_2, 1);
380 PUTBGR24(dst_1, py_1, 1);
381 ENDYUV2RGBLINE(24, 1)
382 LOADCHROMA(0);
383 PUTBGR24(dst_1, py_1, 0);
384 PUTBGR24(dst_2, py_2, 0);
385 ENDYUV2RGBFUNC()
386
387 YUV2RGBFUNC(yuv2rgb_c_16_ordered_dither, uint16_t, 0)
388 const uint8_t *d16 = ff_dither_2x2_8[y & 1];
389 const uint8_t *e16 = ff_dither_2x2_4[y & 1];
390 const uint8_t *f16 = ff_dither_2x2_8[(y & 1)^1];
391
392 #define PUTRGB16(dst, src, i, o) \
393 Y = src[2 * i]; \
394 dst[2 * i] = r[Y + d16[0 + o]] + \
395 g[Y + e16[0 + o]] + \
396 b[Y + f16[0 + o]]; \
397 Y = src[2 * i + 1]; \
398 dst[2 * i + 1] = r[Y + d16[1 + o]] + \
399 g[Y + e16[1 + o]] + \
400 b[Y + f16[1 + o]];
401 LOADCHROMA(0);
402 PUTRGB16(dst_1, py_1, 0, 0);
403 PUTRGB16(dst_2, py_2, 0, 0 + 8);
404
405 LOADCHROMA(1);
406 PUTRGB16(dst_2, py_2, 1, 2 + 8);
407 PUTRGB16(dst_1, py_1, 1, 2);
408
409 LOADCHROMA(2);
410 PUTRGB16(dst_1, py_1, 2, 4);
411 PUTRGB16(dst_2, py_2, 2, 4 + 8);
412
413 LOADCHROMA(3);
414 PUTRGB16(dst_2, py_2, 3, 6 + 8);
415 PUTRGB16(dst_1, py_1, 3, 6);
416 CLOSEYUV2RGBFUNC(8)
417
418 YUV2RGBFUNC(yuv2rgb_c_15_ordered_dither, uint16_t, 0)
419 const uint8_t *d16 = ff_dither_2x2_8[y & 1];
420 const uint8_t *e16 = ff_dither_2x2_8[(y & 1)^1];
421
422 #define PUTRGB15(dst, src, i, o) \
423 Y = src[2 * i]; \
424 dst[2 * i] = r[Y + d16[0 + o]] + \
425 g[Y + d16[1 + o]] + \
426 b[Y + e16[0 + o]]; \
427 Y = src[2 * i + 1]; \
428 dst[2 * i + 1] = r[Y + d16[1 + o]] + \
429 g[Y + d16[0 + o]] + \
430 b[Y + e16[1 + o]];
431 LOADCHROMA(0);
432 PUTRGB15(dst_1, py_1, 0, 0);
433 PUTRGB15(dst_2, py_2, 0, 0 + 8);
434
435 LOADCHROMA(1);
436 PUTRGB15(dst_2, py_2, 1, 2 + 8);
437 PUTRGB15(dst_1, py_1, 1, 2);
438
439 LOADCHROMA(2);
440 PUTRGB15(dst_1, py_1, 2, 4);
441 PUTRGB15(dst_2, py_2, 2, 4 + 8);
442
443 LOADCHROMA(3);
444 PUTRGB15(dst_2, py_2, 3, 6 + 8);
445 PUTRGB15(dst_1, py_1, 3, 6);
446 CLOSEYUV2RGBFUNC(8)
447
448 // r, g, b, dst_1, dst_2
449 YUV2RGBFUNC(yuv2rgb_c_12_ordered_dither, uint16_t, 0)
450 const uint8_t *d16 = ff_dither_4x4_16[y & 3];
451
452 #define PUTRGB12(dst, src, i, o) \
453 Y = src[2 * i]; \
454 dst[2 * i] = r[Y + d16[0 + o]] + \
455 g[Y + d16[0 + o]] + \
456 b[Y + d16[0 + o]]; \
457 Y = src[2 * i + 1]; \
458 dst[2 * i + 1] = r[Y + d16[1 + o]] + \
459 g[Y + d16[1 + o]] + \
460 b[Y + d16[1 + o]];
461
462 LOADCHROMA(0);
463 PUTRGB12(dst_1, py_1, 0, 0);
464 PUTRGB12(dst_2, py_2, 0, 0 + 8);
465
466 LOADCHROMA(1);
467 PUTRGB12(dst_2, py_2, 1, 2 + 8);
468 PUTRGB12(dst_1, py_1, 1, 2);
469
470 LOADCHROMA(2);
471 PUTRGB12(dst_1, py_1, 2, 4);
472 PUTRGB12(dst_2, py_2, 2, 4 + 8);
473
474 LOADCHROMA(3);
475 PUTRGB12(dst_2, py_2, 3, 6 + 8);
476 PUTRGB12(dst_1, py_1, 3, 6);
477 CLOSEYUV2RGBFUNC(8)
478
479 // r, g, b, dst_1, dst_2
480 YUV2RGBFUNC(yuv2rgb_c_8_ordered_dither, uint8_t, 0)
481 const uint8_t *d32 = ff_dither_8x8_32[y & 7];
482 const uint8_t *d64 = ff_dither_8x8_73[y & 7];
483
484 #define PUTRGB8(dst, src, i, o) \
485 Y = src[2 * i]; \
486 dst[2 * i] = r[Y + d32[0 + o]] + \
487 g[Y + d32[0 + o]] + \
488 b[Y + d64[0 + o]]; \
489 Y = src[2 * i + 1]; \
490 dst[2 * i + 1] = r[Y + d32[1 + o]] + \
491 g[Y + d32[1 + o]] + \
492 b[Y + d64[1 + o]];
493
494 LOADCHROMA(0);
495 PUTRGB8(dst_1, py_1, 0, 0);
496 PUTRGB8(dst_2, py_2, 0, 0 + 8);
497
498 LOADCHROMA(1);
499 PUTRGB8(dst_2, py_2, 1, 2 + 8);
500 PUTRGB8(dst_1, py_1, 1, 2);
501
502 LOADCHROMA(2);
503 PUTRGB8(dst_1, py_1, 2, 4);
504 PUTRGB8(dst_2, py_2, 2, 4 + 8);
505
506 LOADCHROMA(3);
507 PUTRGB8(dst_2, py_2, 3, 6 + 8);
508 PUTRGB8(dst_1, py_1, 3, 6);
509 CLOSEYUV2RGBFUNC(8)
510
511 YUV2RGBFUNC(yuv2rgb_c_4_ordered_dither, uint8_t, 0)
512 const uint8_t * d64 = ff_dither_8x8_73[y & 7];
513 const uint8_t *d128 = ff_dither_8x8_220[y & 7];
514 int acc;
515
516 #define PUTRGB4D(dst, src, i, o) \
517 Y = src[2 * i]; \
518 acc = r[Y + d128[0 + o]] + \
519 g[Y + d64[0 + o]] + \
520 b[Y + d128[0 + o]]; \
521 Y = src[2 * i + 1]; \
522 acc |= (r[Y + d128[1 + o]] + \
523 g[Y + d64[1 + o]] + \
524 b[Y + d128[1 + o]]) << 4; \
525 dst[i] = acc;
526
527 LOADCHROMA(0);
528 PUTRGB4D(dst_1, py_1, 0, 0);
529 PUTRGB4D(dst_2, py_2, 0, 0 + 8);
530
531 LOADCHROMA(1);
532 PUTRGB4D(dst_2, py_2, 1, 2 + 8);
533 PUTRGB4D(dst_1, py_1, 1, 2);
534
535 LOADCHROMA(2);
536 PUTRGB4D(dst_1, py_1, 2, 4);
537 PUTRGB4D(dst_2, py_2, 2, 4 + 8);
538
539 LOADCHROMA(3);
540 PUTRGB4D(dst_2, py_2, 3, 6 + 8);
541 PUTRGB4D(dst_1, py_1, 3, 6);
542 CLOSEYUV2RGBFUNC(4)
543
544 YUV2RGBFUNC(yuv2rgb_c_4b_ordered_dither, uint8_t, 0)
545 const uint8_t *d64 = ff_dither_8x8_73[y & 7];
546 const uint8_t *d128 = ff_dither_8x8_220[y & 7];
547
548 #define PUTRGB4DB(dst, src, i, o) \
549 Y = src[2 * i]; \
550 dst[2 * i] = r[Y + d128[0 + o]] + \
551 g[Y + d64[0 + o]] + \
552 b[Y + d128[0 + o]]; \
553 Y = src[2 * i + 1]; \
554 dst[2 * i + 1] = r[Y + d128[1 + o]] + \
555 g[Y + d64[1 + o]] + \
556 b[Y + d128[1 + o]];
557
558 LOADCHROMA(0);
559 PUTRGB4DB(dst_1, py_1, 0, 0);
560 PUTRGB4DB(dst_2, py_2, 0, 0 + 8);
561
562 LOADCHROMA(1);
563 PUTRGB4DB(dst_2, py_2, 1, 2 + 8);
564 PUTRGB4DB(dst_1, py_1, 1, 2);
565
566 LOADCHROMA(2);
567 PUTRGB4DB(dst_1, py_1, 2, 4);
568 PUTRGB4DB(dst_2, py_2, 2, 4 + 8);
569
570 LOADCHROMA(3);
571 PUTRGB4DB(dst_2, py_2, 3, 6 + 8);
572 PUTRGB4DB(dst_1, py_1, 3, 6);
573 CLOSEYUV2RGBFUNC(8)
574
575 YUV2RGBFUNC(yuv2rgb_c_1_ordered_dither, uint8_t, 0)
576 const uint8_t *d128 = ff_dither_8x8_220[y & 7];
577 char out_1 = 0, out_2 = 0;
578 g = c->table_gU[128 + YUVRGB_TABLE_HEADROOM] + c->table_gV[128 + YUVRGB_TABLE_HEADROOM];
579
580 #define PUTRGB1(out, src, i, o) \
581 Y = src[2 * i]; \
582 out += out + g[Y + d128[0 + o]]; \
583 Y = src[2 * i + 1]; \
584 out += out + g[Y + d128[1 + o]];
585
586 PUTRGB1(out_1, py_1, 0, 0);
587 PUTRGB1(out_2, py_2, 0, 0 + 8);
588
589 PUTRGB1(out_2, py_2, 1, 2 + 8);
590 PUTRGB1(out_1, py_1, 1, 2);
591
592 PUTRGB1(out_1, py_1, 2, 4);
593 PUTRGB1(out_2, py_2, 2, 4 + 8);
594
595 PUTRGB1(out_2, py_2, 3, 6 + 8);
596 PUTRGB1(out_1, py_1, 3, 6);
597
598 dst_1[0] = out_1;
599 dst_2[0] = out_2;
600 CLOSEYUV2RGBFUNC(1)
601
ff_yuv2rgb_get_func_ptr(SwsContext * c)602 SwsFunc ff_yuv2rgb_get_func_ptr(SwsContext *c)
603 {
604 SwsFunc t = NULL;
605
606 #if (ARCH_PPC == 1)
607 if (ARCH_PPC)
608 t = ff_yuv2rgb_init_ppc(c);
609 #endif
610 #if (ARCH_X86 == 1)
611 if (ARCH_X86)
612 t = ff_yuv2rgb_init_x86(c);
613 #endif
614
615 if (t)
616 return t;
617
618 av_log(c, AV_LOG_WARNING,
619 "No accelerated colorspace conversion found from %s to %s.\n",
620 av_get_pix_fmt_name(c->srcFormat), av_get_pix_fmt_name(c->dstFormat));
621
622 switch (c->dstFormat) {
623 case AV_PIX_FMT_BGR48BE:
624 case AV_PIX_FMT_BGR48LE:
625 return yuv2rgb_c_bgr48;
626 case AV_PIX_FMT_RGB48BE:
627 case AV_PIX_FMT_RGB48LE:
628 return yuv2rgb_c_48;
629 case AV_PIX_FMT_ARGB:
630 case AV_PIX_FMT_ABGR:
631 if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat))
632 return yuva2argb_c;
633 case AV_PIX_FMT_RGBA:
634 case AV_PIX_FMT_BGRA:
635 return (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat)) ? yuva2rgba_c : yuv2rgb_c_32;
636 case AV_PIX_FMT_RGB24:
637 return yuv2rgb_c_24_rgb;
638 case AV_PIX_FMT_BGR24:
639 return yuv2rgb_c_24_bgr;
640 case AV_PIX_FMT_RGB565:
641 case AV_PIX_FMT_BGR565:
642 return yuv2rgb_c_16_ordered_dither;
643 case AV_PIX_FMT_RGB555:
644 case AV_PIX_FMT_BGR555:
645 return yuv2rgb_c_15_ordered_dither;
646 case AV_PIX_FMT_RGB444:
647 case AV_PIX_FMT_BGR444:
648 return yuv2rgb_c_12_ordered_dither;
649 case AV_PIX_FMT_RGB8:
650 case AV_PIX_FMT_BGR8:
651 return yuv2rgb_c_8_ordered_dither;
652 case AV_PIX_FMT_RGB4:
653 case AV_PIX_FMT_BGR4:
654 return yuv2rgb_c_4_ordered_dither;
655 case AV_PIX_FMT_RGB4_BYTE:
656 case AV_PIX_FMT_BGR4_BYTE:
657 return yuv2rgb_c_4b_ordered_dither;
658 case AV_PIX_FMT_MONOBLACK:
659 return yuv2rgb_c_1_ordered_dither;
660 }
661 return NULL;
662 }
663
fill_table(uint8_t * table[256+2* YUVRGB_TABLE_HEADROOM],const int elemsize,const int64_t inc,void * y_tab)664 static void fill_table(uint8_t* table[256 + 2*YUVRGB_TABLE_HEADROOM], const int elemsize,
665 const int64_t inc, void *y_tab)
666 {
667 int i;
668 uint8_t *y_table = y_tab;
669
670 y_table -= elemsize * (inc >> 9);
671
672 for (i = 0; i < 256 + 2*YUVRGB_TABLE_HEADROOM; i++) {
673 int64_t cb = av_clip(i-YUVRGB_TABLE_HEADROOM, 0, 255)*inc;
674 table[i] = y_table + elemsize * (cb >> 16);
675 }
676 }
677
fill_gv_table(int table[256+2* YUVRGB_TABLE_HEADROOM],const int elemsize,const int64_t inc)678 static void fill_gv_table(int table[256 + 2*YUVRGB_TABLE_HEADROOM], const int elemsize, const int64_t inc)
679 {
680 int i;
681 int off = -(inc >> 9);
682
683 for (i = 0; i < 256 + 2*YUVRGB_TABLE_HEADROOM; i++) {
684 int64_t cb = av_clip(i-YUVRGB_TABLE_HEADROOM, 0, 255)*inc;
685 table[i] = elemsize * (off + (cb >> 16));
686 }
687 }
688
roundToInt16(int64_t f)689 static uint16_t roundToInt16(int64_t f)
690 {
691 int r = (f + (1 << 15)) >> 16;
692
693 if (r < -0x7FFF)
694 return 0x8000;
695 else if (r > 0x7FFF)
696 return 0x7FFF;
697 else
698 return r;
699 }
700
ff_yuv2rgb_c_init_tables(SwsContext * c,const int inv_table[4],int fullRange,int brightness,int contrast,int saturation)701 av_cold int ff_yuv2rgb_c_init_tables(SwsContext *c, const int inv_table[4],
702 int fullRange, int brightness,
703 int contrast, int saturation)
704 {
705 const int isRgb = c->dstFormat == AV_PIX_FMT_RGB32 ||
706 c->dstFormat == AV_PIX_FMT_RGB32_1 ||
707 c->dstFormat == AV_PIX_FMT_BGR24 ||
708 c->dstFormat == AV_PIX_FMT_RGB565BE ||
709 c->dstFormat == AV_PIX_FMT_RGB565LE ||
710 c->dstFormat == AV_PIX_FMT_RGB555BE ||
711 c->dstFormat == AV_PIX_FMT_RGB555LE ||
712 c->dstFormat == AV_PIX_FMT_RGB444BE ||
713 c->dstFormat == AV_PIX_FMT_RGB444LE ||
714 c->dstFormat == AV_PIX_FMT_RGB8 ||
715 c->dstFormat == AV_PIX_FMT_RGB4 ||
716 c->dstFormat == AV_PIX_FMT_RGB4_BYTE ||
717 c->dstFormat == AV_PIX_FMT_MONOBLACK;
718 const int isNotNe = c->dstFormat == AV_PIX_FMT_NE(RGB565LE, RGB565BE) ||
719 c->dstFormat == AV_PIX_FMT_NE(RGB555LE, RGB555BE) ||
720 c->dstFormat == AV_PIX_FMT_NE(RGB444LE, RGB444BE) ||
721 c->dstFormat == AV_PIX_FMT_NE(BGR565LE, BGR565BE) ||
722 c->dstFormat == AV_PIX_FMT_NE(BGR555LE, BGR555BE) ||
723 c->dstFormat == AV_PIX_FMT_NE(BGR444LE, BGR444BE);
724 const int bpp = c->dstFormatBpp;
725 uint8_t *y_table;
726 uint16_t *y_table16;
727 uint32_t *y_table32;
728 int i, base, rbase, gbase, bbase, av_uninit(abase), needAlpha;
729 const int yoffs = fullRange ? 384 : 326;
730
731 int64_t crv = inv_table[0];
732 int64_t cbu = inv_table[1];
733 int64_t cgu = -inv_table[2];
734 int64_t cgv = -inv_table[3];
735 int64_t cy = 1 << 16;
736 int64_t oy = 0;
737 int64_t yb = 0;
738
739 if (!fullRange) {
740 cy = (cy * 255) / 219;
741 oy = 16 << 16;
742 } else {
743 crv = (crv * 224) / 255;
744 cbu = (cbu * 224) / 255;
745 cgu = (cgu * 224) / 255;
746 cgv = (cgv * 224) / 255;
747 }
748
749 cy = (cy * contrast) >> 16;
750 crv = (crv * contrast * saturation) >> 32;
751 cbu = (cbu * contrast * saturation) >> 32;
752 cgu = (cgu * contrast * saturation) >> 32;
753 cgv = (cgv * contrast * saturation) >> 32;
754 oy -= 256 * brightness;
755
756 c->uOffset = LLN(0x0400040004000400);
757 c->vOffset = LLN(0x0400040004000400);
758 c->yCoeff = roundToInt16(cy * 8192) * ULLN(0x0001000100010001);
759 c->vrCoeff = roundToInt16(crv * 8192) * ULLN(0x0001000100010001);
760 c->ubCoeff = roundToInt16(cbu * 8192) * ULLN(0x0001000100010001);
761 c->vgCoeff = roundToInt16(cgv * 8192) * ULLN(0x0001000100010001);
762 c->ugCoeff = roundToInt16(cgu * 8192) * ULLN(0x0001000100010001);
763 c->yOffset = roundToInt16(oy * 8) * ULLN(0x0001000100010001);
764
765 c->yuv2rgb_y_coeff = (int16_t)roundToInt16(cy << 13);
766 c->yuv2rgb_y_offset = (int16_t)roundToInt16(oy << 9);
767 c->yuv2rgb_v2r_coeff = (int16_t)roundToInt16(crv << 13);
768 c->yuv2rgb_v2g_coeff = (int16_t)roundToInt16(cgv << 13);
769 c->yuv2rgb_u2g_coeff = (int16_t)roundToInt16(cgu << 13);
770 c->yuv2rgb_u2b_coeff = (int16_t)roundToInt16(cbu << 13);
771
772 //scale coefficients by cy
773 crv = ((crv << 16) + 0x8000) / FFMAX(cy, 1);
774 cbu = ((cbu << 16) + 0x8000) / FFMAX(cy, 1);
775 cgu = ((cgu << 16) + 0x8000) / FFMAX(cy, 1);
776 cgv = ((cgv << 16) + 0x8000) / FFMAX(cy, 1);
777
778 av_freep(&c->yuvTable);
779
780 switch (bpp) {
781 case 1:
782 c->yuvTable = av_malloc(1024);
783 y_table = c->yuvTable;
784 yb = -(384 << 16) - oy;
785 for (i = 0; i < 1024 - 110; i++) {
786 y_table[i + 110] = av_clip_uint8((yb + 0x8000) >> 16) >> 7;
787 yb += cy;
788 }
789 fill_table(c->table_gU, 1, cgu, y_table + yoffs);
790 fill_gv_table(c->table_gV, 1, cgv);
791 break;
792 case 4:
793 case 4 | 128:
794 rbase = isRgb ? 3 : 0;
795 gbase = 1;
796 bbase = isRgb ? 0 : 3;
797 c->yuvTable = av_malloc(1024 * 3);
798 y_table = c->yuvTable;
799 yb = -(384 << 16) - oy;
800 for (i = 0; i < 1024 - 110; i++) {
801 int yval = av_clip_uint8((yb + 0x8000) >> 16);
802 y_table[i + 110] = (yval >> 7) << rbase;
803 y_table[i + 37 + 1024] = ((yval + 43) / 85) << gbase;
804 y_table[i + 110 + 2048] = (yval >> 7) << bbase;
805 yb += cy;
806 }
807 fill_table(c->table_rV, 1, crv, y_table + yoffs);
808 fill_table(c->table_gU, 1, cgu, y_table + yoffs + 1024);
809 fill_table(c->table_bU, 1, cbu, y_table + yoffs + 2048);
810 fill_gv_table(c->table_gV, 1, cgv);
811 break;
812 case 8:
813 rbase = isRgb ? 5 : 0;
814 gbase = isRgb ? 2 : 3;
815 bbase = isRgb ? 0 : 6;
816 c->yuvTable = av_malloc(1024 * 3);
817 y_table = c->yuvTable;
818 yb = -(384 << 16) - oy;
819 for (i = 0; i < 1024 - 38; i++) {
820 int yval = av_clip_uint8((yb + 0x8000) >> 16);
821 y_table[i + 16] = ((yval + 18) / 36) << rbase;
822 y_table[i + 16 + 1024] = ((yval + 18) / 36) << gbase;
823 y_table[i + 37 + 2048] = ((yval + 43) / 85) << bbase;
824 yb += cy;
825 }
826 fill_table(c->table_rV, 1, crv, y_table + yoffs);
827 fill_table(c->table_gU, 1, cgu, y_table + yoffs + 1024);
828 fill_table(c->table_bU, 1, cbu, y_table + yoffs + 2048);
829 fill_gv_table(c->table_gV, 1, cgv);
830 break;
831 case 12:
832 rbase = isRgb ? 8 : 0;
833 gbase = 4;
834 bbase = isRgb ? 0 : 8;
835 c->yuvTable = av_malloc(1024 * 3 * 2);
836 y_table16 = c->yuvTable;
837 yb = -(384 << 16) - oy;
838 for (i = 0; i < 1024; i++) {
839 uint8_t yval = av_clip_uint8((yb + 0x8000) >> 16);
840 y_table16[i] = (yval >> 4) << rbase;
841 y_table16[i + 1024] = (yval >> 4) << gbase;
842 y_table16[i + 2048] = (yval >> 4) << bbase;
843 yb += cy;
844 }
845 if (isNotNe)
846 for (i = 0; i < 1024 * 3; i++)
847 y_table16[i] = av_bswap16(y_table16[i]);
848 fill_table(c->table_rV, 2, crv, y_table16 + yoffs);
849 fill_table(c->table_gU, 2, cgu, y_table16 + yoffs + 1024);
850 fill_table(c->table_bU, 2, cbu, y_table16 + yoffs + 2048);
851 fill_gv_table(c->table_gV, 2, cgv);
852 break;
853 case 15:
854 case 16:
855 rbase = isRgb ? bpp - 5 : 0;
856 gbase = 5;
857 bbase = isRgb ? 0 : (bpp - 5);
858 c->yuvTable = av_malloc(1024 * 3 * 2);
859 y_table16 = c->yuvTable;
860 yb = -(384 << 16) - oy;
861 for (i = 0; i < 1024; i++) {
862 uint8_t yval = av_clip_uint8((yb + 0x8000) >> 16);
863 y_table16[i] = (yval >> 3) << rbase;
864 y_table16[i + 1024] = (yval >> (18 - bpp)) << gbase;
865 y_table16[i + 2048] = (yval >> 3) << bbase;
866 yb += cy;
867 }
868 if (isNotNe)
869 for (i = 0; i < 1024 * 3; i++)
870 y_table16[i] = av_bswap16(y_table16[i]);
871 fill_table(c->table_rV, 2, crv, y_table16 + yoffs);
872 fill_table(c->table_gU, 2, cgu, y_table16 + yoffs + 1024);
873 fill_table(c->table_bU, 2, cbu, y_table16 + yoffs + 2048);
874 fill_gv_table(c->table_gV, 2, cgv);
875 break;
876 case 24:
877 case 48:
878 c->yuvTable = av_malloc(1024);
879 y_table = c->yuvTable;
880 yb = -(384 << 16) - oy;
881 for (i = 0; i < 1024; i++) {
882 y_table[i] = av_clip_uint8((yb + 0x8000) >> 16);
883 yb += cy;
884 }
885 fill_table(c->table_rV, 1, crv, y_table + yoffs);
886 fill_table(c->table_gU, 1, cgu, y_table + yoffs);
887 fill_table(c->table_bU, 1, cbu, y_table + yoffs);
888 fill_gv_table(c->table_gV, 1, cgv);
889 break;
890 case 32:
891 case 64:
892 base = (c->dstFormat == AV_PIX_FMT_RGB32_1 ||
893 c->dstFormat == AV_PIX_FMT_BGR32_1) ? 8 : 0;
894 rbase = base + (isRgb ? 16 : 0);
895 gbase = base + 8;
896 bbase = base + (isRgb ? 0 : 16);
897 needAlpha = CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat);
898 if (!needAlpha)
899 abase = (base + 24) & 31;
900 c->yuvTable = av_malloc(1024 * 3 * 4);
901 y_table32 = c->yuvTable;
902 yb = -(384 << 16) - oy;
903 for (i = 0; i < 1024; i++) {
904 unsigned yval = av_clip_uint8((yb + 0x8000) >> 16);
905 y_table32[i] = (yval << rbase) +
906 (needAlpha ? 0 : (255u << abase));
907 y_table32[i + 1024] = yval << gbase;
908 y_table32[i + 2048] = yval << bbase;
909 yb += cy;
910 }
911 fill_table(c->table_rV, 4, crv, y_table32 + yoffs);
912 fill_table(c->table_gU, 4, cgu, y_table32 + yoffs + 1024);
913 fill_table(c->table_bU, 4, cbu, y_table32 + yoffs + 2048);
914 fill_gv_table(c->table_gV, 4, cgv);
915 break;
916 default:
917 if(!isPlanar(c->dstFormat) || bpp <= 24)
918 av_log(c, AV_LOG_ERROR, "%ibpp not supported by yuv2rgb\n", bpp);
919 return -1;
920 }
921 return 0;
922 }
923