2 * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
4 * This file is part of Libav.
6 * Libav 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.
11 * Libav 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.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with Libav; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
28 #include "swscale_internal.h"
30 #include "libavutil/intreadwrite.h"
31 #include "libavutil/cpu.h"
32 #include "libavutil/avutil.h"
33 #include "libavutil/mathematics.h"
34 #include "libavutil/bswap.h"
35 #include "libavutil/pixdesc.h"
37 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_1)[8][8] = {
38 { 0, 1, 0, 1, 0, 1, 0, 1,},
39 { 1, 0, 1, 0, 1, 0, 1, 0,},
40 { 0, 1, 0, 1, 0, 1, 0, 1,},
41 { 1, 0, 1, 0, 1, 0, 1, 0,},
42 { 0, 1, 0, 1, 0, 1, 0, 1,},
43 { 1, 0, 1, 0, 1, 0, 1, 0,},
44 { 0, 1, 0, 1, 0, 1, 0, 1,},
45 { 1, 0, 1, 0, 1, 0, 1, 0,},
47 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_3)[8][8] = {
48 { 1, 2, 1, 2, 1, 2, 1, 2,},
49 { 3, 0, 3, 0, 3, 0, 3, 0,},
50 { 1, 2, 1, 2, 1, 2, 1, 2,},
51 { 3, 0, 3, 0, 3, 0, 3, 0,},
52 { 1, 2, 1, 2, 1, 2, 1, 2,},
53 { 3, 0, 3, 0, 3, 0, 3, 0,},
54 { 1, 2, 1, 2, 1, 2, 1, 2,},
55 { 3, 0, 3, 0, 3, 0, 3, 0,},
57 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_64)[8][8] = {
58 { 18, 34, 30, 46, 17, 33, 29, 45,},
59 { 50, 2, 62, 14, 49, 1, 61, 13,},
60 { 26, 42, 22, 38, 25, 41, 21, 37,},
61 { 58, 10, 54, 6, 57, 9, 53, 5,},
62 { 16, 32, 28, 44, 19, 35, 31, 47,},
63 { 48, 0, 60, 12, 51, 3, 63, 15,},
64 { 24, 40, 20, 36, 27, 43, 23, 39,},
65 { 56, 8, 52, 4, 59, 11, 55, 7,},
67 extern const uint8_t dither_8x8_128[8][8];
68 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_256)[8][8] = {
69 { 72, 136, 120, 184, 68, 132, 116, 180,},
70 { 200, 8, 248, 56, 196, 4, 244, 52,},
71 { 104, 168, 88, 152, 100, 164, 84, 148,},
72 { 232, 40, 216, 24, 228, 36, 212, 20,},
73 { 64, 128, 102, 176, 76, 140, 124, 188,},
74 { 192, 0, 240, 48, 204, 12, 252, 60,},
75 { 96, 160, 80, 144, 108, 172, 92, 156,},
76 { 224, 32, 208, 16, 236, 44, 220, 28,},
79 #define RGB2YUV_SHIFT 15
80 #define BY ( (int) (0.114 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
81 #define BV (-(int) (0.081 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
82 #define BU ( (int) (0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
83 #define GY ( (int) (0.587 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
84 #define GV (-(int) (0.419 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
85 #define GU (-(int) (0.331 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
86 #define RY ( (int) (0.299 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
87 #define RV ( (int) (0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
88 #define RU (-(int) (0.169 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
90 static void fillPlane(uint8_t *plane, int stride, int width, int height, int y,
94 uint8_t *ptr = plane + stride * y;
95 for (i = 0; i < height; i++) {
96 memset(ptr, val, width);
101 static void copyPlane(const uint8_t *src, int srcStride,
102 int srcSliceY, int srcSliceH, int width,
103 uint8_t *dst, int dstStride)
105 dst += dstStride * srcSliceY;
106 if (dstStride == srcStride && srcStride > 0) {
107 memcpy(dst, src, srcSliceH * dstStride);
110 for (i = 0; i < srcSliceH; i++) {
111 memcpy(dst, src, width);
118 static int planarToNv12Wrapper(SwsContext *c, const uint8_t *src[],
119 int srcStride[], int srcSliceY,
120 int srcSliceH, uint8_t *dstParam[],
123 uint8_t *dst = dstParam[1] + dstStride[1] * srcSliceY / 2;
125 copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,
126 dstParam[0], dstStride[0]);
128 if (c->dstFormat == PIX_FMT_NV12)
129 interleaveBytes(src[1], src[2], dst, c->srcW / 2, srcSliceH / 2,
130 srcStride[1], srcStride[2], dstStride[0]);
132 interleaveBytes(src[2], src[1], dst, c->srcW / 2, srcSliceH / 2,
133 srcStride[2], srcStride[1], dstStride[0]);
138 static int planarToYuy2Wrapper(SwsContext *c, const uint8_t *src[],
139 int srcStride[], int srcSliceY, int srcSliceH,
140 uint8_t *dstParam[], int dstStride[])
142 uint8_t *dst = dstParam[0] + dstStride[0] * srcSliceY;
144 yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0],
145 srcStride[1], dstStride[0]);
150 static int planarToUyvyWrapper(SwsContext *c, const uint8_t *src[],
151 int srcStride[], int srcSliceY, int srcSliceH,
152 uint8_t *dstParam[], int dstStride[])
154 uint8_t *dst = dstParam[0] + dstStride[0] * srcSliceY;
156 yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0],
157 srcStride[1], dstStride[0]);
162 static int yuv422pToYuy2Wrapper(SwsContext *c, const uint8_t *src[],
163 int srcStride[], int srcSliceY, int srcSliceH,
164 uint8_t *dstParam[], int dstStride[])
166 uint8_t *dst = dstParam[0] + dstStride[0] * srcSliceY;
168 yuv422ptoyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0],
169 srcStride[1], dstStride[0]);
174 static int yuv422pToUyvyWrapper(SwsContext *c, const uint8_t *src[],
175 int srcStride[], int srcSliceY, int srcSliceH,
176 uint8_t *dstParam[], int dstStride[])
178 uint8_t *dst = dstParam[0] + dstStride[0] * srcSliceY;
180 yuv422ptouyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0],
181 srcStride[1], dstStride[0]);
186 static int yuyvToYuv420Wrapper(SwsContext *c, const uint8_t *src[],
187 int srcStride[], int srcSliceY, int srcSliceH,
188 uint8_t *dstParam[], int dstStride[])
190 uint8_t *ydst = dstParam[0] + dstStride[0] * srcSliceY;
191 uint8_t *udst = dstParam[1] + dstStride[1] * srcSliceY / 2;
192 uint8_t *vdst = dstParam[2] + dstStride[2] * srcSliceY / 2;
194 yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0],
195 dstStride[1], srcStride[0]);
198 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
203 static int yuyvToYuv422Wrapper(SwsContext *c, const uint8_t *src[],
204 int srcStride[], int srcSliceY, int srcSliceH,
205 uint8_t *dstParam[], int dstStride[])
207 uint8_t *ydst = dstParam[0] + dstStride[0] * srcSliceY;
208 uint8_t *udst = dstParam[1] + dstStride[1] * srcSliceY;
209 uint8_t *vdst = dstParam[2] + dstStride[2] * srcSliceY;
211 yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0],
212 dstStride[1], srcStride[0]);
217 static int uyvyToYuv420Wrapper(SwsContext *c, const uint8_t *src[],
218 int srcStride[], int srcSliceY, int srcSliceH,
219 uint8_t *dstParam[], int dstStride[])
221 uint8_t *ydst = dstParam[0] + dstStride[0] * srcSliceY;
222 uint8_t *udst = dstParam[1] + dstStride[1] * srcSliceY / 2;
223 uint8_t *vdst = dstParam[2] + dstStride[2] * srcSliceY / 2;
225 uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0],
226 dstStride[1], srcStride[0]);
229 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
234 static int uyvyToYuv422Wrapper(SwsContext *c, const uint8_t *src[],
235 int srcStride[], int srcSliceY, int srcSliceH,
236 uint8_t *dstParam[], int dstStride[])
238 uint8_t *ydst = dstParam[0] + dstStride[0] * srcSliceY;
239 uint8_t *udst = dstParam[1] + dstStride[1] * srcSliceY;
240 uint8_t *vdst = dstParam[2] + dstStride[2] * srcSliceY;
242 uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0],
243 dstStride[1], srcStride[0]);
248 static void gray8aToPacked32(const uint8_t *src, uint8_t *dst, int num_pixels,
249 const uint8_t *palette)
252 for (i = 0; i < num_pixels; i++)
253 ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i << 1]] | (src[(i << 1) + 1] << 24);
256 static void gray8aToPacked32_1(const uint8_t *src, uint8_t *dst, int num_pixels,
257 const uint8_t *palette)
261 for (i = 0; i < num_pixels; i++)
262 ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i << 1]] | src[(i << 1) + 1];
265 static void gray8aToPacked24(const uint8_t *src, uint8_t *dst, int num_pixels,
266 const uint8_t *palette)
270 for (i = 0; i < num_pixels; i++) {
272 dst[0] = palette[src[i << 1] * 4 + 0];
273 dst[1] = palette[src[i << 1] * 4 + 1];
274 dst[2] = palette[src[i << 1] * 4 + 2];
279 static int packed_16bpc_bswap(SwsContext *c, const uint8_t *src[],
280 int srcStride[], int srcSliceY, int srcSliceH,
281 uint8_t *dst[], int dstStride[])
284 int srcstr = srcStride[0] >> 1;
285 int dststr = dstStride[0] >> 1;
286 uint16_t *dstPtr = (uint16_t *) dst[0];
287 const uint16_t *srcPtr = (const uint16_t *) src[0];
289 for (i = 0; i < srcSliceH; i++) {
290 for (j = 0; j < srcstr; j++) {
291 dstPtr[j] = av_bswap16(srcPtr[j]);
300 static int palToRgbWrapper(SwsContext *c, const uint8_t *src[], int srcStride[],
301 int srcSliceY, int srcSliceH, uint8_t *dst[],
304 const enum PixelFormat srcFormat = c->srcFormat;
305 const enum PixelFormat dstFormat = c->dstFormat;
306 void (*conv)(const uint8_t *src, uint8_t *dst, int num_pixels,
307 const uint8_t *palette) = NULL;
309 uint8_t *dstPtr = dst[0] + dstStride[0] * srcSliceY;
310 const uint8_t *srcPtr = src[0];
312 if (srcFormat == PIX_FMT_Y400A) {
314 case PIX_FMT_RGB32 : conv = gray8aToPacked32; break;
315 case PIX_FMT_BGR32 : conv = gray8aToPacked32; break;
316 case PIX_FMT_BGR32_1: conv = gray8aToPacked32_1; break;
317 case PIX_FMT_RGB32_1: conv = gray8aToPacked32_1; break;
318 case PIX_FMT_RGB24 : conv = gray8aToPacked24; break;
319 case PIX_FMT_BGR24 : conv = gray8aToPacked24; break;
321 } else if (usePal(srcFormat)) {
323 case PIX_FMT_RGB32 : conv = sws_convertPalette8ToPacked32; break;
324 case PIX_FMT_BGR32 : conv = sws_convertPalette8ToPacked32; break;
325 case PIX_FMT_BGR32_1: conv = sws_convertPalette8ToPacked32; break;
326 case PIX_FMT_RGB32_1: conv = sws_convertPalette8ToPacked32; break;
327 case PIX_FMT_RGB24 : conv = sws_convertPalette8ToPacked24; break;
328 case PIX_FMT_BGR24 : conv = sws_convertPalette8ToPacked24; break;
333 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
334 sws_format_name(srcFormat), sws_format_name(dstFormat));
336 for (i = 0; i < srcSliceH; i++) {
337 conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
338 srcPtr += srcStride[0];
339 dstPtr += dstStride[0];
346 #define isRGBA32(x) ( \
347 (x) == PIX_FMT_ARGB \
348 || (x) == PIX_FMT_RGBA \
349 || (x) == PIX_FMT_BGRA \
350 || (x) == PIX_FMT_ABGR \
353 /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
354 static int rgbToRgbWrapper(SwsContext *c, const uint8_t *src[], int srcStride[],
355 int srcSliceY, int srcSliceH, uint8_t *dst[],
358 const enum PixelFormat srcFormat = c->srcFormat;
359 const enum PixelFormat dstFormat = c->dstFormat;
360 const int srcBpp = (c->srcFormatBpp + 7) >> 3;
361 const int dstBpp = (c->dstFormatBpp + 7) >> 3;
362 const int srcId = c->srcFormatBpp >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
363 const int dstId = c->dstFormatBpp >> 2;
364 void (*conv)(const uint8_t *src, uint8_t *dst, int src_size) = NULL;
366 #define CONV_IS(src, dst) (srcFormat == PIX_FMT_##src && dstFormat == PIX_FMT_##dst)
368 if (isRGBA32(srcFormat) && isRGBA32(dstFormat)) {
369 if ( CONV_IS(ABGR, RGBA)
370 || CONV_IS(ARGB, BGRA)
371 || CONV_IS(BGRA, ARGB)
372 || CONV_IS(RGBA, ABGR)) conv = shuffle_bytes_3210;
373 else if (CONV_IS(ABGR, ARGB)
374 || CONV_IS(ARGB, ABGR)) conv = shuffle_bytes_0321;
375 else if (CONV_IS(ABGR, BGRA)
376 || CONV_IS(ARGB, RGBA)) conv = shuffle_bytes_1230;
377 else if (CONV_IS(BGRA, RGBA)
378 || CONV_IS(RGBA, BGRA)) conv = shuffle_bytes_2103;
379 else if (CONV_IS(BGRA, ABGR)
380 || CONV_IS(RGBA, ARGB)) conv = shuffle_bytes_3012;
383 if ((isBGRinInt(srcFormat) && isBGRinInt(dstFormat)) ||
384 (isRGBinInt(srcFormat) && isRGBinInt(dstFormat))) {
385 switch (srcId | (dstId << 4)) {
386 case 0x34: conv = rgb16to15; break;
387 case 0x36: conv = rgb24to15; break;
388 case 0x38: conv = rgb32to15; break;
389 case 0x43: conv = rgb15to16; break;
390 case 0x46: conv = rgb24to16; break;
391 case 0x48: conv = rgb32to16; break;
392 case 0x63: conv = rgb15to24; break;
393 case 0x64: conv = rgb16to24; break;
394 case 0x68: conv = rgb32to24; break;
395 case 0x83: conv = rgb15to32; break;
396 case 0x84: conv = rgb16to32; break;
397 case 0x86: conv = rgb24to32; break;
399 } else if ((isBGRinInt(srcFormat) && isRGBinInt(dstFormat)) ||
400 (isRGBinInt(srcFormat) && isBGRinInt(dstFormat))) {
401 switch (srcId | (dstId << 4)) {
402 case 0x33: conv = rgb15tobgr15; break;
403 case 0x34: conv = rgb16tobgr15; break;
404 case 0x36: conv = rgb24tobgr15; break;
405 case 0x38: conv = rgb32tobgr15; break;
406 case 0x43: conv = rgb15tobgr16; break;
407 case 0x44: conv = rgb16tobgr16; break;
408 case 0x46: conv = rgb24tobgr16; break;
409 case 0x48: conv = rgb32tobgr16; break;
410 case 0x63: conv = rgb15tobgr24; break;
411 case 0x64: conv = rgb16tobgr24; break;
412 case 0x66: conv = rgb24tobgr24; break;
413 case 0x68: conv = rgb32tobgr24; break;
414 case 0x83: conv = rgb15tobgr32; break;
415 case 0x84: conv = rgb16tobgr32; break;
416 case 0x86: conv = rgb24tobgr32; break;
421 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
422 sws_format_name(srcFormat), sws_format_name(dstFormat));
424 const uint8_t *srcPtr = src[0];
425 uint8_t *dstPtr = dst[0];
426 if ((srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1) &&
427 !isRGBA32(dstFormat))
428 srcPtr += ALT32_CORR;
430 if ((dstFormat == PIX_FMT_RGB32_1 || dstFormat == PIX_FMT_BGR32_1) &&
431 !isRGBA32(srcFormat))
432 dstPtr += ALT32_CORR;
434 if (dstStride[0] * srcBpp == srcStride[0] * dstBpp && srcStride[0] > 0 &&
435 !(srcStride[0] % srcBpp))
436 conv(srcPtr, dstPtr + dstStride[0] * srcSliceY,
437 srcSliceH * srcStride[0]);
440 dstPtr += dstStride[0] * srcSliceY;
442 for (i = 0; i < srcSliceH; i++) {
443 conv(srcPtr, dstPtr, c->srcW * srcBpp);
444 srcPtr += srcStride[0];
445 dstPtr += dstStride[0];
452 static int bgr24ToYv12Wrapper(SwsContext *c, const uint8_t *src[],
453 int srcStride[], int srcSliceY, int srcSliceH,
454 uint8_t *dst[], int dstStride[])
458 dst[0] + srcSliceY * dstStride[0],
459 dst[1] + (srcSliceY >> 1) * dstStride[1],
460 dst[2] + (srcSliceY >> 1) * dstStride[2],
462 dstStride[0], dstStride[1], srcStride[0]);
464 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
468 static int yvu9ToYv12Wrapper(SwsContext *c, const uint8_t *src[],
469 int srcStride[], int srcSliceY, int srcSliceH,
470 uint8_t *dst[], int dstStride[])
472 copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,
473 dst[0], dstStride[0]);
475 planar2x(src[1], dst[1] + dstStride[1] * (srcSliceY >> 1), c->chrSrcW,
476 srcSliceH >> 2, srcStride[1], dstStride[1]);
477 planar2x(src[2], dst[2] + dstStride[2] * (srcSliceY >> 1), c->chrSrcW,
478 srcSliceH >> 2, srcStride[2], dstStride[2]);
480 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
484 /* unscaled copy like stuff (assumes nearly identical formats) */
485 static int packedCopyWrapper(SwsContext *c, const uint8_t *src[],
486 int srcStride[], int srcSliceY, int srcSliceH,
487 uint8_t *dst[], int dstStride[])
489 if (dstStride[0] == srcStride[0] && srcStride[0] > 0)
490 memcpy(dst[0] + dstStride[0] * srcSliceY, src[0], srcSliceH * dstStride[0]);
493 const uint8_t *srcPtr = src[0];
494 uint8_t *dstPtr = dst[0] + dstStride[0] * srcSliceY;
497 /* universal length finder */
498 while (length + c->srcW <= FFABS(dstStride[0]) &&
499 length + c->srcW <= FFABS(srcStride[0]))
503 for (i = 0; i < srcSliceH; i++) {
504 memcpy(dstPtr, srcPtr, length);
505 srcPtr += srcStride[0];
506 dstPtr += dstStride[0];
512 #define clip9(x) av_clip_uintp2(x, 9)
513 #define clip10(x) av_clip_uintp2(x, 10)
514 #define DITHER_COPY(dst, dstStride, wfunc, src, srcStride, rfunc, dithers, shift, clip) \
515 for (i = 0; i < height; i++) { \
516 const uint8_t *dither = dithers[i & 7]; \
517 for (j = 0; j < length - 7; j += 8) { \
518 wfunc(&dst[j + 0], clip((rfunc(&src[j + 0]) + dither[0]) >> shift)); \
519 wfunc(&dst[j + 1], clip((rfunc(&src[j + 1]) + dither[1]) >> shift)); \
520 wfunc(&dst[j + 2], clip((rfunc(&src[j + 2]) + dither[2]) >> shift)); \
521 wfunc(&dst[j + 3], clip((rfunc(&src[j + 3]) + dither[3]) >> shift)); \
522 wfunc(&dst[j + 4], clip((rfunc(&src[j + 4]) + dither[4]) >> shift)); \
523 wfunc(&dst[j + 5], clip((rfunc(&src[j + 5]) + dither[5]) >> shift)); \
524 wfunc(&dst[j + 6], clip((rfunc(&src[j + 6]) + dither[6]) >> shift)); \
525 wfunc(&dst[j + 7], clip((rfunc(&src[j + 7]) + dither[7]) >> shift)); \
527 for (; j < length; j++) \
528 wfunc(&dst[j], (rfunc(&src[j]) + dither[j & 7]) >> shift); \
533 static int planarCopyWrapper(SwsContext *c, const uint8_t *src[],
534 int srcStride[], int srcSliceY, int srcSliceH,
535 uint8_t *dst[], int dstStride[])
538 for (plane = 0; plane < 4; plane++) {
539 int length = (plane == 0 || plane == 3) ? c->srcW : -((-c->srcW ) >> c->chrDstHSubSample);
540 int y = (plane == 0 || plane == 3) ? srcSliceY: -((-srcSliceY) >> c->chrDstVSubSample);
541 int height = (plane == 0 || plane == 3) ? srcSliceH: -((-srcSliceH) >> c->chrDstVSubSample);
542 const uint8_t *srcPtr = src[plane];
543 uint8_t *dstPtr = dst[plane] + dstStride[plane] * y;
547 // ignore palette for GRAY8
548 if (plane == 1 && !dst[2]) continue;
549 if (!src[plane] || (plane == 1 && !src[2])) {
550 if (is16BPS(c->dstFormat))
552 fillPlane(dst[plane], dstStride[plane], length, height, y,
553 (plane == 3) ? 255 : 128);
555 if (is9_OR_10BPS(c->srcFormat)) {
556 const int src_depth = av_pix_fmt_descriptors[c->srcFormat].comp[plane].depth_minus1 + 1;
557 const int dst_depth = av_pix_fmt_descriptors[c->dstFormat].comp[plane].depth_minus1 + 1;
558 const uint16_t *srcPtr2 = (const uint16_t *) srcPtr;
560 if (is16BPS(c->dstFormat)) {
561 uint16_t *dstPtr2 = (uint16_t *) dstPtr;
562 #define COPY9_OR_10TO16(rfunc, wfunc) \
563 for (i = 0; i < height; i++) { \
564 for (j = 0; j < length; j++) { \
565 int srcpx = rfunc(&srcPtr2[j]); \
566 wfunc(&dstPtr2[j], (srcpx << (16 - src_depth)) | (srcpx >> (2 * src_depth - 16))); \
568 dstPtr2 += dstStride[plane] / 2; \
569 srcPtr2 += srcStride[plane] / 2; \
571 if (isBE(c->dstFormat)) {
572 if (isBE(c->srcFormat)) {
573 COPY9_OR_10TO16(AV_RB16, AV_WB16);
575 COPY9_OR_10TO16(AV_RL16, AV_WB16);
578 if (isBE(c->srcFormat)) {
579 COPY9_OR_10TO16(AV_RB16, AV_WL16);
581 COPY9_OR_10TO16(AV_RL16, AV_WL16);
584 } else if (is9_OR_10BPS(c->dstFormat)) {
585 uint16_t *dstPtr2 = (uint16_t *) dstPtr;
586 #define COPY9_OR_10TO9_OR_10(loop) \
587 for (i = 0; i < height; i++) { \
588 for (j = 0; j < length; j++) { \
591 dstPtr2 += dstStride[plane] / 2; \
592 srcPtr2 += srcStride[plane] / 2; \
594 #define COPY9_OR_10TO9_OR_10_2(rfunc, wfunc) \
595 if (dst_depth > src_depth) { \
596 COPY9_OR_10TO9_OR_10(int srcpx = rfunc(&srcPtr2[j]); \
597 wfunc(&dstPtr2[j], (srcpx << 1) | (srcpx >> 9))); \
598 } else if (dst_depth < src_depth) { \
599 DITHER_COPY(dstPtr2, dstStride[plane] / 2, wfunc, \
600 srcPtr2, srcStride[plane] / 2, rfunc, \
601 dither_8x8_1, 1, clip9); \
603 COPY9_OR_10TO9_OR_10(wfunc(&dstPtr2[j], rfunc(&srcPtr2[j]))); \
605 if (isBE(c->dstFormat)) {
606 if (isBE(c->srcFormat)) {
607 COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WB16);
609 COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WB16);
612 if (isBE(c->srcFormat)) {
613 COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WL16);
615 COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WL16);
619 #define W8(a, b) { *(a) = (b); }
620 #define COPY9_OR_10TO8(rfunc) \
621 if (src_depth == 9) { \
622 DITHER_COPY(dstPtr, dstStride[plane], W8, \
623 srcPtr2, srcStride[plane] / 2, rfunc, \
624 dither_8x8_1, 1, av_clip_uint8); \
626 DITHER_COPY(dstPtr, dstStride[plane], W8, \
627 srcPtr2, srcStride[plane] / 2, rfunc, \
628 dither_8x8_3, 2, av_clip_uint8); \
630 if (isBE(c->srcFormat)) {
631 COPY9_OR_10TO8(AV_RB16);
633 COPY9_OR_10TO8(AV_RL16);
636 } else if (is9_OR_10BPS(c->dstFormat)) {
637 const int dst_depth = av_pix_fmt_descriptors[c->dstFormat].comp[plane].depth_minus1 + 1;
638 uint16_t *dstPtr2 = (uint16_t *) dstPtr;
640 if (is16BPS(c->srcFormat)) {
641 const uint16_t *srcPtr2 = (const uint16_t *) srcPtr;
642 #define COPY16TO9_OR_10(rfunc, wfunc) \
643 if (dst_depth == 9) { \
644 DITHER_COPY(dstPtr2, dstStride[plane] / 2, wfunc, \
645 srcPtr2, srcStride[plane] / 2, rfunc, \
646 dither_8x8_128, 7, clip9); \
648 DITHER_COPY(dstPtr2, dstStride[plane] / 2, wfunc, \
649 srcPtr2, srcStride[plane] / 2, rfunc, \
650 dither_8x8_64, 6, clip10); \
652 if (isBE(c->dstFormat)) {
653 if (isBE(c->srcFormat)) {
654 COPY16TO9_OR_10(AV_RB16, AV_WB16);
656 COPY16TO9_OR_10(AV_RL16, AV_WB16);
659 if (isBE(c->srcFormat)) {
660 COPY16TO9_OR_10(AV_RB16, AV_WL16);
662 COPY16TO9_OR_10(AV_RL16, AV_WL16);
666 #define COPY8TO9_OR_10(wfunc) \
667 for (i = 0; i < height; i++) { \
668 for (j = 0; j < length; j++) { \
669 const int srcpx = srcPtr[j]; \
670 wfunc(&dstPtr2[j], (srcpx << (dst_depth - 8)) | (srcpx >> (16 - dst_depth))); \
672 dstPtr2 += dstStride[plane] / 2; \
673 srcPtr += srcStride[plane]; \
675 if (isBE(c->dstFormat)) {
676 COPY8TO9_OR_10(AV_WB16);
678 COPY8TO9_OR_10(AV_WL16);
681 } else if (is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {
682 const uint16_t *srcPtr2 = (const uint16_t *) srcPtr;
683 #define COPY16TO8(rfunc) \
684 DITHER_COPY(dstPtr, dstStride[plane], W8, \
685 srcPtr2, srcStride[plane] / 2, rfunc, \
686 dither_8x8_256, 8, av_clip_uint8);
687 if (isBE(c->srcFormat)) {
692 } else if (!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {
693 for (i = 0; i < height; i++) {
694 for (j = 0; j < length; j++) {
695 dstPtr[ j << 1 ] = srcPtr[j];
696 dstPtr[(j << 1) + 1] = srcPtr[j];
698 srcPtr += srcStride[plane];
699 dstPtr += dstStride[plane];
701 } else if (is16BPS(c->srcFormat) && is16BPS(c->dstFormat) &&
702 isBE(c->srcFormat) != isBE(c->dstFormat)) {
704 for (i = 0; i < height; i++) {
705 for (j = 0; j < length; j++)
706 ((uint16_t *) dstPtr)[j] = av_bswap16(((const uint16_t *) srcPtr)[j]);
707 srcPtr += srcStride[plane];
708 dstPtr += dstStride[plane];
710 } else if (dstStride[plane] == srcStride[plane] &&
711 srcStride[plane] > 0 && srcStride[plane] == length) {
712 memcpy(dst[plane] + dstStride[plane] * y, src[plane],
713 height * dstStride[plane]);
715 if (is16BPS(c->srcFormat) && is16BPS(c->dstFormat))
717 for (i = 0; i < height; i++) {
718 memcpy(dstPtr, srcPtr, length);
719 srcPtr += srcStride[plane];
720 dstPtr += dstStride[plane];
729 #define IS_DIFFERENT_ENDIANESS(src_fmt, dst_fmt, pix_fmt) \
730 ((src_fmt == pix_fmt ## BE && dst_fmt == pix_fmt ## LE) || \
731 (src_fmt == pix_fmt ## LE && dst_fmt == pix_fmt ## BE))
734 void ff_get_unscaled_swscale(SwsContext *c)
736 const enum PixelFormat srcFormat = c->srcFormat;
737 const enum PixelFormat dstFormat = c->dstFormat;
738 const int flags = c->flags;
739 const int dstH = c->dstH;
742 needsDither = isAnyRGB(dstFormat) &&
743 c->dstFormatBpp < 24 &&
744 (c->dstFormatBpp < c->srcFormatBpp || (!isAnyRGB(srcFormat)));
747 if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) &&
748 (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {
749 c->swScale = planarToNv12Wrapper;
752 if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUV422P ||
753 srcFormat == PIX_FMT_YUVA420P) && isAnyRGB(dstFormat) &&
754 !(flags & SWS_ACCURATE_RND) && !(dstH & 1)) {
755 c->swScale = ff_yuv2rgb_get_func_ptr(c);
758 if (srcFormat == PIX_FMT_YUV410P &&
759 (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P) &&
760 !(flags & SWS_BITEXACT)) {
761 c->swScale = yvu9ToYv12Wrapper;
765 if (srcFormat == PIX_FMT_BGR24 &&
766 (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P) &&
767 !(flags & SWS_ACCURATE_RND))
768 c->swScale = bgr24ToYv12Wrapper;
770 /* RGB/BGR -> RGB/BGR (no dither needed forms) */
771 if ( isAnyRGB(srcFormat)
772 && isAnyRGB(dstFormat)
773 && srcFormat != PIX_FMT_BGR8 && dstFormat != PIX_FMT_BGR8
774 && srcFormat != PIX_FMT_RGB8 && dstFormat != PIX_FMT_RGB8
775 && srcFormat != PIX_FMT_BGR4 && dstFormat != PIX_FMT_BGR4
776 && srcFormat != PIX_FMT_RGB4 && dstFormat != PIX_FMT_RGB4
777 && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
778 && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
779 && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
780 && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
781 && srcFormat != PIX_FMT_RGB48LE && dstFormat != PIX_FMT_RGB48LE
782 && srcFormat != PIX_FMT_RGB48BE && dstFormat != PIX_FMT_RGB48BE
783 && srcFormat != PIX_FMT_BGR48LE && dstFormat != PIX_FMT_BGR48LE
784 && srcFormat != PIX_FMT_BGR48BE && dstFormat != PIX_FMT_BGR48BE
785 && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
786 c->swScale= rgbToRgbWrapper;
788 /* bswap 16 bits per pixel/component packed formats */
789 if (IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, PIX_FMT_BGR444) ||
790 IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, PIX_FMT_BGR48) ||
791 IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, PIX_FMT_BGR555) ||
792 IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, PIX_FMT_BGR565) ||
793 IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, PIX_FMT_GRAY16) ||
794 IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, PIX_FMT_RGB444) ||
795 IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, PIX_FMT_RGB48) ||
796 IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, PIX_FMT_RGB555) ||
797 IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, PIX_FMT_RGB565))
798 c->swScale = packed_16bpc_bswap;
800 if ((usePal(srcFormat) && (
801 dstFormat == PIX_FMT_RGB32 ||
802 dstFormat == PIX_FMT_RGB32_1 ||
803 dstFormat == PIX_FMT_RGB24 ||
804 dstFormat == PIX_FMT_BGR32 ||
805 dstFormat == PIX_FMT_BGR32_1 ||
806 dstFormat == PIX_FMT_BGR24)))
807 c->swScale = palToRgbWrapper;
809 if (srcFormat == PIX_FMT_YUV422P) {
810 if (dstFormat == PIX_FMT_YUYV422)
811 c->swScale = yuv422pToYuy2Wrapper;
812 else if (dstFormat == PIX_FMT_UYVY422)
813 c->swScale = yuv422pToUyvyWrapper;
816 /* LQ converters if -sws 0 or -sws 4*/
817 if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {
819 if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {
820 if (dstFormat == PIX_FMT_YUYV422)
821 c->swScale = planarToYuy2Wrapper;
822 else if (dstFormat == PIX_FMT_UYVY422)
823 c->swScale = planarToUyvyWrapper;
826 if (srcFormat == PIX_FMT_YUYV422 &&
827 (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
828 c->swScale = yuyvToYuv420Wrapper;
829 if (srcFormat == PIX_FMT_UYVY422 &&
830 (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
831 c->swScale = uyvyToYuv420Wrapper;
832 if (srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)
833 c->swScale = yuyvToYuv422Wrapper;
834 if (srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
835 c->swScale = uyvyToYuv422Wrapper;
838 if ( srcFormat == dstFormat ||
839 (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P) ||
840 (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P) ||
841 (isPlanarYUV(srcFormat) && isGray(dstFormat)) ||
842 (isPlanarYUV(dstFormat) && isGray(srcFormat)) ||
843 (isGray(dstFormat) && isGray(srcFormat)) ||
844 (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat) &&
845 c->chrDstHSubSample == c->chrSrcHSubSample &&
846 c->chrDstVSubSample == c->chrSrcVSubSample &&
847 dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21 &&
848 srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))
850 if (isPacked(c->srcFormat))
851 c->swScale = packedCopyWrapper;
852 else /* Planar YUV or gray */
853 c->swScale = planarCopyWrapper;
857 ff_bfin_get_unscaled_swscale(c);
859 ff_swscale_get_unscaled_altivec(c);
862 static void reset_ptr(const uint8_t *src[], int format)
864 if (!isALPHA(format))
866 if (!isPlanar(format)) {
867 src[3] = src[2] = NULL;
874 static int check_image_pointers(uint8_t *data[4], enum PixelFormat pix_fmt,
875 const int linesizes[4])
877 const AVPixFmtDescriptor *desc = &av_pix_fmt_descriptors[pix_fmt];
880 for (i = 0; i < 4; i++) {
881 int plane = desc->comp[i].plane;
882 if (!data[plane] || !linesizes[plane])
890 * swscale wrapper, so we don't need to export the SwsContext.
891 * Assumes planar YUV to be in YUV order instead of YVU.
893 int attribute_align_arg sws_scale(struct SwsContext *c,
894 const uint8_t * const srcSlice[],
895 const int srcStride[], int srcSliceY,
896 int srcSliceH, uint8_t *const dst[],
897 const int dstStride[])
900 const uint8_t *src2[4] = { srcSlice[0], srcSlice[1], srcSlice[2], srcSlice[3] };
901 uint8_t *dst2[4] = { dst[0], dst[1], dst[2], dst[3] };
903 // do not mess up sliceDir if we have a "trailing" 0-size slice
907 if (!check_image_pointers(srcSlice, c->srcFormat, srcStride)) {
908 av_log(c, AV_LOG_ERROR, "bad src image pointers\n");
911 if (!check_image_pointers(dst, c->dstFormat, dstStride)) {
912 av_log(c, AV_LOG_ERROR, "bad dst image pointers\n");
916 if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
917 av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
920 if (c->sliceDir == 0) {
921 if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
924 if (usePal(c->srcFormat)) {
925 for (i = 0; i < 256; i++) {
926 int p, r, g, b, y, u, v;
927 if (c->srcFormat == PIX_FMT_PAL8) {
928 p = ((const uint32_t *)(srcSlice[1]))[i];
929 r = (p >> 16) & 0xFF;
932 } else if (c->srcFormat == PIX_FMT_RGB8) {
934 g = ((i >> 2) & 7) * 36;
936 } else if (c->srcFormat == PIX_FMT_BGR8) {
938 g = ((i >> 3) & 7) * 36;
940 } else if (c->srcFormat == PIX_FMT_RGB4_BYTE) {
941 r = ( i >> 3 ) * 255;
942 g = ((i >> 1) & 3) * 85;
944 } else if (c->srcFormat == PIX_FMT_GRAY8 ||
945 c->srcFormat == PIX_FMT_Y400A) {
948 assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
949 b = ( i >> 3 ) * 255;
950 g = ((i >> 1) & 3) * 85;
953 y = av_clip_uint8((RY * r + GY * g + BY * b + ( 33 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);
954 u = av_clip_uint8((RU * r + GU * g + BU * b + (257 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);
955 v = av_clip_uint8((RV * r + GV * g + BV * b + (257 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);
956 c->pal_yuv[i] = y + (u << 8) + (v << 16);
958 switch (c->dstFormat) {
963 c->pal_rgb[i] = r + (g << 8) + (b << 16);
965 case PIX_FMT_BGR32_1:
969 c->pal_rgb[i] = (r + (g << 8) + (b << 16)) << 8;
971 case PIX_FMT_RGB32_1:
975 c->pal_rgb[i] = (b + (g << 8) + (r << 16)) << 8;
982 c->pal_rgb[i] = b + (g << 8) + (r << 16);
987 // copy strides, so they can safely be modified
988 if (c->sliceDir == 1) {
989 // slices go from top to bottom
990 int srcStride2[4] = { srcStride[0], srcStride[1], srcStride[2],
992 int dstStride2[4] = { dstStride[0], dstStride[1], dstStride[2],
995 reset_ptr(src2, c->srcFormat);
996 reset_ptr((const uint8_t **) dst2, c->dstFormat);
998 /* reset slice direction at end of frame */
999 if (srcSliceY + srcSliceH == c->srcH)
1002 return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2,
1005 // slices go from bottom to top => we flip the image internally
1006 int srcStride2[4] = { -srcStride[0], -srcStride[1], -srcStride[2],
1008 int dstStride2[4] = { -dstStride[0], -dstStride[1], -dstStride[2],
1011 src2[0] += (srcSliceH - 1) * srcStride[0];
1012 if (!usePal(c->srcFormat))
1013 src2[1] += ((srcSliceH >> c->chrSrcVSubSample) - 1) * srcStride[1];
1014 src2[2] += ((srcSliceH >> c->chrSrcVSubSample) - 1) * srcStride[2];
1015 src2[3] += (srcSliceH - 1) * srcStride[3];
1016 dst2[0] += ( c->dstH - 1) * dstStride[0];
1017 dst2[1] += ((c->dstH >> c->chrDstVSubSample) - 1) * dstStride[1];
1018 dst2[2] += ((c->dstH >> c->chrDstVSubSample) - 1) * dstStride[2];
1019 dst2[3] += ( c->dstH - 1) * dstStride[3];
1021 reset_ptr(src2, c->srcFormat);
1022 reset_ptr((const uint8_t **) dst2, c->dstFormat);
1024 /* reset slice direction at end of frame */
1028 return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH,
1029 srcSliceH, dst2, dstStride2);
1033 /* Convert the palette to the same packed 32-bit format as the palette */
1034 void sws_convertPalette8ToPacked32(const uint8_t *src, uint8_t *dst,
1035 int num_pixels, const uint8_t *palette)
1039 for (i = 0; i < num_pixels; i++)
1040 ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i]];
1043 /* Palette format: ABCD -> dst format: ABC */
1044 void sws_convertPalette8ToPacked24(const uint8_t *src, uint8_t *dst,
1045 int num_pixels, const uint8_t *palette)
1049 for (i = 0; i < num_pixels; i++) {
1051 dst[0] = palette[src[i] * 4 + 0];
1052 dst[1] = palette[src[i] * 4 + 1];
1053 dst[2] = palette[src[i] * 4 + 2];