2 * Copyright (C) 2001-2011 Michael Niedermayer <michaelni@gmx.at>
4 * This file is part of FFmpeg.
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.
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.
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
21 #ifndef SWSCALE_SWSCALE_INTERNAL_H
22 #define SWSCALE_SWSCALE_INTERNAL_H
27 #include "libavutil/avassert.h"
28 #include "libavutil/avutil.h"
29 #include "libavutil/common.h"
30 #include "libavutil/intreadwrite.h"
31 #include "libavutil/log.h"
32 #include "libavutil/pixfmt.h"
33 #include "libavutil/pixdesc.h"
34 #include "libavutil/ppc/util_altivec.h"
36 #define STR(s) AV_TOSTRING(s) // AV_STRINGIFY is too long
38 #define YUVRGB_TABLE_HEADROOM 512
39 #define YUVRGB_TABLE_LUMA_HEADROOM 512
41 #define MAX_FILTER_SIZE SWS_MAX_FILTER_SIZE
46 #define ALT32_CORR (-1)
61 #define RETCODE_USE_CASCADE -12345
65 typedef enum SwsDither {
75 typedef enum SwsAlphaBlend {
76 SWS_ALPHA_BLEND_NONE = 0,
77 SWS_ALPHA_BLEND_UNIFORM,
78 SWS_ALPHA_BLEND_CHECKERBOARD,
82 typedef int (*SwsFunc)(struct SwsContext *context, const uint8_t *src[],
83 int srcStride[], int srcSliceY, int srcSliceH,
84 uint8_t *dst[], int dstStride[]);
87 * Write one line of horizontally scaled data to planar output
88 * without any additional vertical scaling (or point-scaling).
90 * @param src scaled source data, 15 bits for 8-10-bit output,
91 * 19 bits for 16-bit output (in int32_t)
92 * @param dest pointer to the output plane. For >8-bit
93 * output, this is in uint16_t
94 * @param dstW width of destination in pixels
95 * @param dither ordered dither array of type int16_t and size 8
96 * @param offset Dither offset
98 typedef void (*yuv2planar1_fn)(const int16_t *src, uint8_t *dest, int dstW,
99 const uint8_t *dither, int offset);
102 * Write one line of horizontally scaled data to planar output
103 * with multi-point vertical scaling between input pixels.
105 * @param filter vertical luma/alpha scaling coefficients, 12 bits [0,4096]
106 * @param src scaled luma (Y) or alpha (A) source data, 15 bits for
107 * 8-10-bit output, 19 bits for 16-bit output (in int32_t)
108 * @param filterSize number of vertical input lines to scale
109 * @param dest pointer to output plane. For >8-bit
110 * output, this is in uint16_t
111 * @param dstW width of destination pixels
112 * @param offset Dither offset
114 typedef void (*yuv2planarX_fn)(const int16_t *filter, int filterSize,
115 const int16_t **src, uint8_t *dest, int dstW,
116 const uint8_t *dither, int offset);
119 * Write one line of horizontally scaled chroma to interleaved output
120 * with multi-point vertical scaling between input pixels.
122 * @param dstFormat destination pixel format
123 * @param chrDither ordered dither array of type uint8_t and size 8
124 * @param chrFilter vertical chroma scaling coefficients, 12 bits [0,4096]
125 * @param chrUSrc scaled chroma (U) source data, 15 bits for 8-10-bit
126 * output, 19 bits for 16-bit output (in int32_t)
127 * @param chrVSrc scaled chroma (V) source data, 15 bits for 8-10-bit
128 * output, 19 bits for 16-bit output (in int32_t)
129 * @param chrFilterSize number of vertical chroma input lines to scale
130 * @param dest pointer to the output plane. For >8-bit
131 * output, this is in uint16_t
132 * @param dstW width of chroma planes
134 typedef void (*yuv2interleavedX_fn)(enum AVPixelFormat dstFormat,
135 const uint8_t *chrDither,
136 const int16_t *chrFilter,
138 const int16_t **chrUSrc,
139 const int16_t **chrVSrc,
140 uint8_t *dest, int dstW);
143 * Write one line of horizontally scaled Y/U/V/A to packed-pixel YUV/RGB
144 * output without any additional vertical scaling (or point-scaling). Note
145 * that this function may do chroma scaling, see the "uvalpha" argument.
147 * @param c SWS scaling context
148 * @param lumSrc scaled luma (Y) source data, 15 bits for 8-10-bit output,
149 * 19 bits for 16-bit output (in int32_t)
150 * @param chrUSrc scaled chroma (U) source data, 15 bits for 8-10-bit output,
151 * 19 bits for 16-bit output (in int32_t)
152 * @param chrVSrc scaled chroma (V) source data, 15 bits for 8-10-bit output,
153 * 19 bits for 16-bit output (in int32_t)
154 * @param alpSrc scaled alpha (A) source data, 15 bits for 8-10-bit output,
155 * 19 bits for 16-bit output (in int32_t)
156 * @param dest pointer to the output plane. For 16-bit output, this is
158 * @param dstW width of lumSrc and alpSrc in pixels, number of pixels
159 * to write into dest[]
160 * @param uvalpha chroma scaling coefficient for the second line of chroma
161 * pixels, either 2048 or 0. If 0, one chroma input is used
162 * for 2 output pixels (or if the SWS_FLAG_FULL_CHR_INT flag
163 * is set, it generates 1 output pixel). If 2048, two chroma
164 * input pixels should be averaged for 2 output pixels (this
165 * only happens if SWS_FLAG_FULL_CHR_INT is not set)
166 * @param y vertical line number for this output. This does not need
167 * to be used to calculate the offset in the destination,
168 * but can be used to generate comfort noise using dithering
169 * for some output formats.
171 typedef void (*yuv2packed1_fn)(struct SwsContext *c, const int16_t *lumSrc,
172 const int16_t *chrUSrc[2],
173 const int16_t *chrVSrc[2],
174 const int16_t *alpSrc, uint8_t *dest,
175 int dstW, int uvalpha, int y);
177 * Write one line of horizontally scaled Y/U/V/A to packed-pixel YUV/RGB
178 * output by doing bilinear scaling between two input lines.
180 * @param c SWS scaling context
181 * @param lumSrc scaled luma (Y) source data, 15 bits for 8-10-bit output,
182 * 19 bits for 16-bit output (in int32_t)
183 * @param chrUSrc scaled chroma (U) source data, 15 bits for 8-10-bit output,
184 * 19 bits for 16-bit output (in int32_t)
185 * @param chrVSrc scaled chroma (V) source data, 15 bits for 8-10-bit output,
186 * 19 bits for 16-bit output (in int32_t)
187 * @param alpSrc scaled alpha (A) source data, 15 bits for 8-10-bit output,
188 * 19 bits for 16-bit output (in int32_t)
189 * @param dest pointer to the output plane. For 16-bit output, this is
191 * @param dstW width of lumSrc and alpSrc in pixels, number of pixels
192 * to write into dest[]
193 * @param yalpha luma/alpha scaling coefficients for the second input line.
194 * The first line's coefficients can be calculated by using
196 * @param uvalpha chroma scaling coefficient for the second input line. The
197 * first line's coefficients can be calculated by using
199 * @param y vertical line number for this output. This does not need
200 * to be used to calculate the offset in the destination,
201 * but can be used to generate comfort noise using dithering
202 * for some output formats.
204 typedef void (*yuv2packed2_fn)(struct SwsContext *c, const int16_t *lumSrc[2],
205 const int16_t *chrUSrc[2],
206 const int16_t *chrVSrc[2],
207 const int16_t *alpSrc[2],
209 int dstW, int yalpha, int uvalpha, int y);
211 * Write one line of horizontally scaled Y/U/V/A to packed-pixel YUV/RGB
212 * output by doing multi-point vertical scaling between input pixels.
214 * @param c SWS scaling context
215 * @param lumFilter vertical luma/alpha scaling coefficients, 12 bits [0,4096]
216 * @param lumSrc scaled luma (Y) source data, 15 bits for 8-10-bit output,
217 * 19 bits for 16-bit output (in int32_t)
218 * @param lumFilterSize number of vertical luma/alpha input lines to scale
219 * @param chrFilter vertical chroma scaling coefficients, 12 bits [0,4096]
220 * @param chrUSrc scaled chroma (U) source data, 15 bits for 8-10-bit output,
221 * 19 bits for 16-bit output (in int32_t)
222 * @param chrVSrc scaled chroma (V) source data, 15 bits for 8-10-bit output,
223 * 19 bits for 16-bit output (in int32_t)
224 * @param chrFilterSize number of vertical chroma input lines to scale
225 * @param alpSrc scaled alpha (A) source data, 15 bits for 8-10-bit output,
226 * 19 bits for 16-bit output (in int32_t)
227 * @param dest pointer to the output plane. For 16-bit output, this is
229 * @param dstW width of lumSrc and alpSrc in pixels, number of pixels
230 * to write into dest[]
231 * @param y vertical line number for this output. This does not need
232 * to be used to calculate the offset in the destination,
233 * but can be used to generate comfort noise using dithering
234 * or some output formats.
236 typedef void (*yuv2packedX_fn)(struct SwsContext *c, const int16_t *lumFilter,
237 const int16_t **lumSrc, int lumFilterSize,
238 const int16_t *chrFilter,
239 const int16_t **chrUSrc,
240 const int16_t **chrVSrc, int chrFilterSize,
241 const int16_t **alpSrc, uint8_t *dest,
245 * Write one line of horizontally scaled Y/U/V/A to YUV/RGB
246 * output by doing multi-point vertical scaling between input pixels.
248 * @param c SWS scaling context
249 * @param lumFilter vertical luma/alpha scaling coefficients, 12 bits [0,4096]
250 * @param lumSrc scaled luma (Y) source data, 15 bits for 8-10-bit output,
251 * 19 bits for 16-bit output (in int32_t)
252 * @param lumFilterSize number of vertical luma/alpha input lines to scale
253 * @param chrFilter vertical chroma scaling coefficients, 12 bits [0,4096]
254 * @param chrUSrc scaled chroma (U) source data, 15 bits for 8-10-bit output,
255 * 19 bits for 16-bit output (in int32_t)
256 * @param chrVSrc scaled chroma (V) source data, 15 bits for 8-10-bit output,
257 * 19 bits for 16-bit output (in int32_t)
258 * @param chrFilterSize number of vertical chroma input lines to scale
259 * @param alpSrc scaled alpha (A) source data, 15 bits for 8-10-bit output,
260 * 19 bits for 16-bit output (in int32_t)
261 * @param dest pointer to the output planes. For 16-bit output, this is
263 * @param dstW width of lumSrc and alpSrc in pixels, number of pixels
264 * to write into dest[]
265 * @param y vertical line number for this output. This does not need
266 * to be used to calculate the offset in the destination,
267 * but can be used to generate comfort noise using dithering
268 * or some output formats.
270 typedef void (*yuv2anyX_fn)(struct SwsContext *c, const int16_t *lumFilter,
271 const int16_t **lumSrc, int lumFilterSize,
272 const int16_t *chrFilter,
273 const int16_t **chrUSrc,
274 const int16_t **chrVSrc, int chrFilterSize,
275 const int16_t **alpSrc, uint8_t **dest,
279 struct SwsFilterDescriptor;
281 /* This struct should be aligned on at least a 32-byte boundary. */
282 typedef struct SwsContext {
284 * info on struct for av_log
286 const AVClass *av_class;
289 * Note that src, dst, srcStride, dstStride will be copied in the
290 * sws_scale() wrapper so they can be freely modified here.
293 int srcW; ///< Width of source luma/alpha planes.
294 int srcH; ///< Height of source luma/alpha planes.
295 int dstH; ///< Height of destination luma/alpha planes.
296 int chrSrcW; ///< Width of source chroma planes.
297 int chrSrcH; ///< Height of source chroma planes.
298 int chrDstW; ///< Width of destination chroma planes.
299 int chrDstH; ///< Height of destination chroma planes.
300 int lumXInc, chrXInc;
301 int lumYInc, chrYInc;
302 enum AVPixelFormat dstFormat; ///< Destination pixel format.
303 enum AVPixelFormat srcFormat; ///< Source pixel format.
304 int dstFormatBpp; ///< Number of bits per pixel of the destination pixel format.
305 int srcFormatBpp; ///< Number of bits per pixel of the source pixel format.
307 int chrSrcHSubSample; ///< Binary logarithm of horizontal subsampling factor between luma/alpha and chroma planes in source image.
308 int chrSrcVSubSample; ///< Binary logarithm of vertical subsampling factor between luma/alpha and chroma planes in source image.
309 int chrDstHSubSample; ///< Binary logarithm of horizontal subsampling factor between luma/alpha and chroma planes in destination image.
310 int chrDstVSubSample; ///< Binary logarithm of vertical subsampling factor between luma/alpha and chroma planes in destination image.
311 int vChrDrop; ///< Binary logarithm of extra vertical subsampling factor in source image chroma planes specified by user.
312 int sliceDir; ///< Direction that slices are fed to the scaler (1 = top-to-bottom, -1 = bottom-to-top).
313 double param[2]; ///< Input parameters for scaling algorithms that need them.
315 /* The cascaded_* fields allow spliting a scaler task into multiple
316 * sequential steps, this is for example used to limit the maximum
317 * downscaling factor that needs to be supported in one scaler.
319 struct SwsContext *cascaded_context[3];
320 int cascaded_tmpStride[4];
321 uint8_t *cascaded_tmp[4];
322 int cascaded1_tmpStride[4];
323 uint8_t *cascaded1_tmp[4];
324 int cascaded_mainindex;
328 int is_internal_gamma;
335 struct SwsSlice *slice;
336 struct SwsFilterDescriptor *desc;
338 uint32_t pal_yuv[256];
339 uint32_t pal_rgb[256];
341 float uint2float_lut[256];
344 * @name Scaled horizontal lines ring buffer.
345 * The horizontal scaler keeps just enough scaled lines in a ring buffer
346 * so they may be passed to the vertical scaler. The pointers to the
347 * allocated buffers for each line are duplicated in sequence in the ring
348 * buffer to simplify indexing and avoid wrapping around between lines
349 * inside the vertical scaler code. The wrapping is done before the
350 * vertical scaler is called.
353 int lastInLumBuf; ///< Last scaled horizontal luma/alpha line from source in the ring buffer.
354 int lastInChrBuf; ///< Last scaled horizontal chroma line from source in the ring buffer.
357 uint8_t *formatConvBuffer;
361 * @name Horizontal and vertical filters.
362 * To better understand the following fields, here is a pseudo-code of
363 * their usage in filtering a horizontal line:
365 * for (i = 0; i < width; i++) {
367 * for (j = 0; j < filterSize; j++)
368 * dst[i] += src[ filterPos[i] + j ] * filter[ filterSize * i + j ];
369 * dst[i] >>= FRAC_BITS; // The actual implementation is fixed-point.
374 int16_t *hLumFilter; ///< Array of horizontal filter coefficients for luma/alpha planes.
375 int16_t *hChrFilter; ///< Array of horizontal filter coefficients for chroma planes.
376 int16_t *vLumFilter; ///< Array of vertical filter coefficients for luma/alpha planes.
377 int16_t *vChrFilter; ///< Array of vertical filter coefficients for chroma planes.
378 int32_t *hLumFilterPos; ///< Array of horizontal filter starting positions for each dst[i] for luma/alpha planes.
379 int32_t *hChrFilterPos; ///< Array of horizontal filter starting positions for each dst[i] for chroma planes.
380 int32_t *vLumFilterPos; ///< Array of vertical filter starting positions for each dst[i] for luma/alpha planes.
381 int32_t *vChrFilterPos; ///< Array of vertical filter starting positions for each dst[i] for chroma planes.
382 int hLumFilterSize; ///< Horizontal filter size for luma/alpha pixels.
383 int hChrFilterSize; ///< Horizontal filter size for chroma pixels.
384 int vLumFilterSize; ///< Vertical filter size for luma/alpha pixels.
385 int vChrFilterSize; ///< Vertical filter size for chroma pixels.
388 int lumMmxextFilterCodeSize; ///< Runtime-generated MMXEXT horizontal fast bilinear scaler code size for luma/alpha planes.
389 int chrMmxextFilterCodeSize; ///< Runtime-generated MMXEXT horizontal fast bilinear scaler code size for chroma planes.
390 uint8_t *lumMmxextFilterCode; ///< Runtime-generated MMXEXT horizontal fast bilinear scaler code for luma/alpha planes.
391 uint8_t *chrMmxextFilterCode; ///< Runtime-generated MMXEXT horizontal fast bilinear scaler code for chroma planes.
394 int warned_unuseable_bilinear;
396 int dstY; ///< Last destination vertical line output from last slice.
397 int flags; ///< Flags passed by the user to select scaler algorithm, optimizations, subsampling, etc...
398 void *yuvTable; // pointer to the yuv->rgb table start so it can be freed()
399 // alignment ensures the offset can be added in a single
400 // instruction on e.g. ARM
401 DECLARE_ALIGNED(16, int, table_gV)[256 + 2*YUVRGB_TABLE_HEADROOM];
402 uint8_t *table_rV[256 + 2*YUVRGB_TABLE_HEADROOM];
403 uint8_t *table_gU[256 + 2*YUVRGB_TABLE_HEADROOM];
404 uint8_t *table_bU[256 + 2*YUVRGB_TABLE_HEADROOM];
405 DECLARE_ALIGNED(16, int32_t, input_rgb2yuv_table)[16+40*4]; // This table can contain both C and SIMD formatted values, the C vales are always at the XY_IDX points
415 #define RGB2YUV_SHIFT 15
417 int *dither_error[4];
420 int contrast, brightness, saturation; // for sws_getColorspaceDetails
421 int srcColorspaceTable[4];
422 int dstColorspaceTable[4];
423 int srcRange; ///< 0 = MPG YUV range, 1 = JPG YUV range (source image).
424 int dstRange; ///< 0 = MPG YUV range, 1 = JPG YUV range (destination image).
433 int yuv2rgb_y_offset;
435 int yuv2rgb_v2r_coeff;
436 int yuv2rgb_v2g_coeff;
437 int yuv2rgb_u2g_coeff;
438 int yuv2rgb_u2b_coeff;
440 #define RED_DITHER "0*8"
441 #define GREEN_DITHER "1*8"
442 #define BLUE_DITHER "2*8"
443 #define Y_COEFF "3*8"
444 #define VR_COEFF "4*8"
445 #define UB_COEFF "5*8"
446 #define VG_COEFF "6*8"
447 #define UG_COEFF "7*8"
448 #define Y_OFFSET "8*8"
449 #define U_OFFSET "9*8"
450 #define V_OFFSET "10*8"
451 #define LUM_MMX_FILTER_OFFSET "11*8"
452 #define CHR_MMX_FILTER_OFFSET "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)
453 #define DSTW_OFFSET "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*2"
454 #define ESP_OFFSET "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*2+8"
455 #define VROUNDER_OFFSET "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*2+16"
456 #define U_TEMP "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*2+24"
457 #define V_TEMP "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*2+32"
458 #define Y_TEMP "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*2+40"
459 #define ALP_MMX_FILTER_OFFSET "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*2+48"
460 #define UV_OFF_PX "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*3+48"
461 #define UV_OFF_BYTE "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*3+56"
462 #define DITHER16 "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*3+64"
463 #define DITHER32 "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*3+80"
464 #define DITHER32_INT (11*8+4*4*MAX_FILTER_SIZE*3+80) // value equal to above, used for checking that the struct hasn't been changed by mistake
466 DECLARE_ALIGNED(8, uint64_t, redDither);
467 DECLARE_ALIGNED(8, uint64_t, greenDither);
468 DECLARE_ALIGNED(8, uint64_t, blueDither);
470 DECLARE_ALIGNED(8, uint64_t, yCoeff);
471 DECLARE_ALIGNED(8, uint64_t, vrCoeff);
472 DECLARE_ALIGNED(8, uint64_t, ubCoeff);
473 DECLARE_ALIGNED(8, uint64_t, vgCoeff);
474 DECLARE_ALIGNED(8, uint64_t, ugCoeff);
475 DECLARE_ALIGNED(8, uint64_t, yOffset);
476 DECLARE_ALIGNED(8, uint64_t, uOffset);
477 DECLARE_ALIGNED(8, uint64_t, vOffset);
478 int32_t lumMmxFilter[4 * MAX_FILTER_SIZE];
479 int32_t chrMmxFilter[4 * MAX_FILTER_SIZE];
480 int dstW; ///< Width of destination luma/alpha planes.
481 DECLARE_ALIGNED(8, uint64_t, esp);
482 DECLARE_ALIGNED(8, uint64_t, vRounder);
483 DECLARE_ALIGNED(8, uint64_t, u_temp);
484 DECLARE_ALIGNED(8, uint64_t, v_temp);
485 DECLARE_ALIGNED(8, uint64_t, y_temp);
486 int32_t alpMmxFilter[4 * MAX_FILTER_SIZE];
487 // alignment of these values is not necessary, but merely here
488 // to maintain the same offset across x8632 and x86-64. Once we
489 // use proper offset macros in the asm, they can be removed.
490 DECLARE_ALIGNED(8, ptrdiff_t, uv_off); ///< offset (in pixels) between u and v planes
491 DECLARE_ALIGNED(8, ptrdiff_t, uv_offx2); ///< offset (in bytes) between u and v planes
492 DECLARE_ALIGNED(8, uint16_t, dither16)[8];
493 DECLARE_ALIGNED(8, uint32_t, dither32)[8];
495 const uint8_t *chrDither8, *lumDither8;
498 vector signed short CY;
499 vector signed short CRV;
500 vector signed short CBU;
501 vector signed short CGU;
502 vector signed short CGV;
503 vector signed short OY;
504 vector unsigned short CSHIFT;
505 vector signed short *vYCoeffsBank, *vCCoeffsBank;
510 /* pre defined color-spaces gamma */
511 #define XYZ_GAMMA (2.6f)
512 #define RGB_GAMMA (2.2f)
515 int16_t *xyzgammainv;
516 int16_t *rgbgammainv;
517 int16_t xyz2rgb_matrix[3][4];
518 int16_t rgb2xyz_matrix[3][4];
520 /* function pointers for swscale() */
521 yuv2planar1_fn yuv2plane1;
522 yuv2planarX_fn yuv2planeX;
523 yuv2interleavedX_fn yuv2nv12cX;
524 yuv2packed1_fn yuv2packed1;
525 yuv2packed2_fn yuv2packed2;
526 yuv2packedX_fn yuv2packedX;
527 yuv2anyX_fn yuv2anyX;
529 /// Unscaled conversion of luma plane to YV12 for horizontal scaler.
530 void (*lumToYV12)(uint8_t *dst, const uint8_t *src, const uint8_t *src2, const uint8_t *src3,
531 int width, uint32_t *pal);
532 /// Unscaled conversion of alpha plane to YV12 for horizontal scaler.
533 void (*alpToYV12)(uint8_t *dst, const uint8_t *src, const uint8_t *src2, const uint8_t *src3,
534 int width, uint32_t *pal);
535 /// Unscaled conversion of chroma planes to YV12 for horizontal scaler.
536 void (*chrToYV12)(uint8_t *dstU, uint8_t *dstV,
537 const uint8_t *src1, const uint8_t *src2, const uint8_t *src3,
538 int width, uint32_t *pal);
541 * Functions to read planar input, such as planar RGB, and convert
542 * internally to Y/UV/A.
545 void (*readLumPlanar)(uint8_t *dst, const uint8_t *src[4], int width, int32_t *rgb2yuv);
546 void (*readChrPlanar)(uint8_t *dstU, uint8_t *dstV, const uint8_t *src[4],
547 int width, int32_t *rgb2yuv);
548 void (*readAlpPlanar)(uint8_t *dst, const uint8_t *src[4], int width, int32_t *rgb2yuv);
552 * Scale one horizontal line of input data using a bilinear filter
553 * to produce one line of output data. Compared to SwsContext->hScale(),
554 * please take note of the following caveats when using these:
555 * - Scaling is done using only 7 bits instead of 14-bit coefficients.
556 * - You can use no more than 5 input pixels to produce 4 output
557 * pixels. Therefore, this filter should not be used for downscaling
558 * by more than ~20% in width (because that equals more than 5/4th
559 * downscaling and thus more than 5 pixels input per 4 pixels output).
560 * - In general, bilinear filters create artifacts during downscaling
561 * (even when <20%), because one output pixel will span more than one
562 * input pixel, and thus some pixels will need edges of both neighbor
563 * pixels to interpolate the output pixel. Since you can use at most
564 * two input pixels per output pixel in bilinear scaling, this is
565 * impossible and thus downscaling by any size will create artifacts.
566 * To enable this type of scaling, set SWS_FLAG_FAST_BILINEAR
567 * in SwsContext->flags.
570 void (*hyscale_fast)(struct SwsContext *c,
571 int16_t *dst, int dstWidth,
572 const uint8_t *src, int srcW, int xInc);
573 void (*hcscale_fast)(struct SwsContext *c,
574 int16_t *dst1, int16_t *dst2, int dstWidth,
575 const uint8_t *src1, const uint8_t *src2,
580 * Scale one horizontal line of input data using a filter over the input
581 * lines, to produce one (differently sized) line of output data.
583 * @param dst pointer to destination buffer for horizontally scaled
584 * data. If the number of bits per component of one
585 * destination pixel (SwsContext->dstBpc) is <= 10, data
586 * will be 15 bpc in 16 bits (int16_t) width. Else (i.e.
587 * SwsContext->dstBpc == 16), data will be 19bpc in
588 * 32 bits (int32_t) width.
589 * @param dstW width of destination image
590 * @param src pointer to source data to be scaled. If the number of
591 * bits per component of a source pixel (SwsContext->srcBpc)
592 * is 8, this is 8bpc in 8 bits (uint8_t) width. Else
593 * (i.e. SwsContext->dstBpc > 8), this is native depth
594 * in 16 bits (uint16_t) width. In other words, for 9-bit
595 * YUV input, this is 9bpc, for 10-bit YUV input, this is
596 * 10bpc, and for 16-bit RGB or YUV, this is 16bpc.
597 * @param filter filter coefficients to be used per output pixel for
598 * scaling. This contains 14bpp filtering coefficients.
599 * Guaranteed to contain dstW * filterSize entries.
600 * @param filterPos position of the first input pixel to be used for
601 * each output pixel during scaling. Guaranteed to
602 * contain dstW entries.
603 * @param filterSize the number of input coefficients to be used (and
604 * thus the number of input pixels to be used) for
605 * creating a single output pixel. Is aligned to 4
606 * (and input coefficients thus padded with zeroes)
607 * to simplify creating SIMD code.
610 void (*hyScale)(struct SwsContext *c, int16_t *dst, int dstW,
611 const uint8_t *src, const int16_t *filter,
612 const int32_t *filterPos, int filterSize);
613 void (*hcScale)(struct SwsContext *c, int16_t *dst, int dstW,
614 const uint8_t *src, const int16_t *filter,
615 const int32_t *filterPos, int filterSize);
618 /// Color range conversion function for luma plane if needed.
619 void (*lumConvertRange)(int16_t *dst, int width);
620 /// Color range conversion function for chroma planes if needed.
621 void (*chrConvertRange)(int16_t *dst1, int16_t *dst2, int width);
623 int needs_hcscale; ///< Set if there are chroma planes to be converted.
627 SwsAlphaBlend alphablend;
629 //FIXME check init (where 0)
631 SwsFunc ff_yuv2rgb_get_func_ptr(SwsContext *c);
632 int ff_yuv2rgb_c_init_tables(SwsContext *c, const int inv_table[4],
633 int fullRange, int brightness,
634 int contrast, int saturation);
635 void ff_yuv2rgb_init_tables_ppc(SwsContext *c, const int inv_table[4],
636 int brightness, int contrast, int saturation);
638 void ff_updateMMXDitherTables(SwsContext *c, int dstY);
640 av_cold void ff_sws_init_range_convert(SwsContext *c);
642 SwsFunc ff_yuv2rgb_init_x86(SwsContext *c);
643 SwsFunc ff_yuv2rgb_init_ppc(SwsContext *c);
645 static av_always_inline int is16BPS(enum AVPixelFormat pix_fmt)
647 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
649 return desc->comp[0].depth == 16;
652 static av_always_inline int is32BPS(enum AVPixelFormat pix_fmt)
654 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
656 return desc->comp[0].depth == 32;
659 static av_always_inline int isNBPS(enum AVPixelFormat pix_fmt)
661 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
663 return desc->comp[0].depth >= 9 && desc->comp[0].depth <= 14;
666 static av_always_inline int isBE(enum AVPixelFormat pix_fmt)
668 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
670 return desc->flags & AV_PIX_FMT_FLAG_BE;
673 static av_always_inline int isYUV(enum AVPixelFormat pix_fmt)
675 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
677 return !(desc->flags & AV_PIX_FMT_FLAG_RGB) && desc->nb_components >= 2;
680 static av_always_inline int isPlanarYUV(enum AVPixelFormat pix_fmt)
682 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
684 return ((desc->flags & AV_PIX_FMT_FLAG_PLANAR) && isYUV(pix_fmt));
688 * Identity semi-planar YUV formats. Specifically, those are YUV formats
689 * where the second and third components (U & V) are on the same plane.
691 static av_always_inline int isSemiPlanarYUV(enum AVPixelFormat pix_fmt)
693 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
695 return (isPlanarYUV(pix_fmt) && desc->comp[1].plane == desc->comp[2].plane);
698 static av_always_inline int isRGB(enum AVPixelFormat pix_fmt)
700 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
702 return (desc->flags & AV_PIX_FMT_FLAG_RGB);
705 static av_always_inline int isGray(enum AVPixelFormat pix_fmt)
707 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
709 return !(desc->flags & AV_PIX_FMT_FLAG_PAL) &&
710 !(desc->flags & AV_PIX_FMT_FLAG_HWACCEL) &&
711 desc->nb_components <= 2 &&
712 pix_fmt != AV_PIX_FMT_MONOBLACK &&
713 pix_fmt != AV_PIX_FMT_MONOWHITE;
716 static av_always_inline int isRGBinInt(enum AVPixelFormat pix_fmt)
718 return pix_fmt == AV_PIX_FMT_RGB48BE ||
719 pix_fmt == AV_PIX_FMT_RGB48LE ||
720 pix_fmt == AV_PIX_FMT_RGB32 ||
721 pix_fmt == AV_PIX_FMT_RGB32_1 ||
722 pix_fmt == AV_PIX_FMT_RGB24 ||
723 pix_fmt == AV_PIX_FMT_RGB565BE ||
724 pix_fmt == AV_PIX_FMT_RGB565LE ||
725 pix_fmt == AV_PIX_FMT_RGB555BE ||
726 pix_fmt == AV_PIX_FMT_RGB555LE ||
727 pix_fmt == AV_PIX_FMT_RGB444BE ||
728 pix_fmt == AV_PIX_FMT_RGB444LE ||
729 pix_fmt == AV_PIX_FMT_RGB8 ||
730 pix_fmt == AV_PIX_FMT_RGB4 ||
731 pix_fmt == AV_PIX_FMT_RGB4_BYTE ||
732 pix_fmt == AV_PIX_FMT_RGBA64BE ||
733 pix_fmt == AV_PIX_FMT_RGBA64LE ||
734 pix_fmt == AV_PIX_FMT_MONOBLACK ||
735 pix_fmt == AV_PIX_FMT_MONOWHITE;
738 static av_always_inline int isBGRinInt(enum AVPixelFormat pix_fmt)
740 return pix_fmt == AV_PIX_FMT_BGR48BE ||
741 pix_fmt == AV_PIX_FMT_BGR48LE ||
742 pix_fmt == AV_PIX_FMT_BGR32 ||
743 pix_fmt == AV_PIX_FMT_BGR32_1 ||
744 pix_fmt == AV_PIX_FMT_BGR24 ||
745 pix_fmt == AV_PIX_FMT_BGR565BE ||
746 pix_fmt == AV_PIX_FMT_BGR565LE ||
747 pix_fmt == AV_PIX_FMT_BGR555BE ||
748 pix_fmt == AV_PIX_FMT_BGR555LE ||
749 pix_fmt == AV_PIX_FMT_BGR444BE ||
750 pix_fmt == AV_PIX_FMT_BGR444LE ||
751 pix_fmt == AV_PIX_FMT_BGR8 ||
752 pix_fmt == AV_PIX_FMT_BGR4 ||
753 pix_fmt == AV_PIX_FMT_BGR4_BYTE ||
754 pix_fmt == AV_PIX_FMT_BGRA64BE ||
755 pix_fmt == AV_PIX_FMT_BGRA64LE ||
756 pix_fmt == AV_PIX_FMT_MONOBLACK ||
757 pix_fmt == AV_PIX_FMT_MONOWHITE;
760 static av_always_inline int isBayer(enum AVPixelFormat pix_fmt)
762 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
764 return !!(desc->flags & AV_PIX_FMT_FLAG_BAYER);
767 static av_always_inline int isBayer16BPS(enum AVPixelFormat pix_fmt)
769 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
771 return desc->comp[1].depth == 8;
774 static av_always_inline int isAnyRGB(enum AVPixelFormat pix_fmt)
776 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
778 return (desc->flags & AV_PIX_FMT_FLAG_RGB) ||
779 pix_fmt == AV_PIX_FMT_MONOBLACK || pix_fmt == AV_PIX_FMT_MONOWHITE;
782 static av_always_inline int isFloat(enum AVPixelFormat pix_fmt)
784 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
786 return desc->flags & AV_PIX_FMT_FLAG_FLOAT;
789 static av_always_inline int isALPHA(enum AVPixelFormat pix_fmt)
791 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
793 if (pix_fmt == AV_PIX_FMT_PAL8)
795 return desc->flags & AV_PIX_FMT_FLAG_ALPHA;
798 static av_always_inline int isPacked(enum AVPixelFormat pix_fmt)
800 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
802 return (desc->nb_components >= 2 && !(desc->flags & AV_PIX_FMT_FLAG_PLANAR)) ||
803 pix_fmt == AV_PIX_FMT_PAL8 ||
804 pix_fmt == AV_PIX_FMT_MONOBLACK || pix_fmt == AV_PIX_FMT_MONOWHITE;
807 static av_always_inline int isPlanar(enum AVPixelFormat pix_fmt)
809 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
811 return (desc->nb_components >= 2 && (desc->flags & AV_PIX_FMT_FLAG_PLANAR));
814 static av_always_inline int isPackedRGB(enum AVPixelFormat pix_fmt)
816 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
818 return ((desc->flags & (AV_PIX_FMT_FLAG_PLANAR | AV_PIX_FMT_FLAG_RGB)) == AV_PIX_FMT_FLAG_RGB);
821 static av_always_inline int isPlanarRGB(enum AVPixelFormat pix_fmt)
823 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
825 return ((desc->flags & (AV_PIX_FMT_FLAG_PLANAR | AV_PIX_FMT_FLAG_RGB)) ==
826 (AV_PIX_FMT_FLAG_PLANAR | AV_PIX_FMT_FLAG_RGB));
829 static av_always_inline int usePal(enum AVPixelFormat pix_fmt)
832 case AV_PIX_FMT_PAL8:
833 case AV_PIX_FMT_BGR4_BYTE:
834 case AV_PIX_FMT_BGR8:
835 case AV_PIX_FMT_GRAY8:
836 case AV_PIX_FMT_RGB4_BYTE:
837 case AV_PIX_FMT_RGB8:
844 extern const uint64_t ff_dither4[2];
845 extern const uint64_t ff_dither8[2];
847 extern const uint8_t ff_dither_2x2_4[3][8];
848 extern const uint8_t ff_dither_2x2_8[3][8];
849 extern const uint8_t ff_dither_4x4_16[5][8];
850 extern const uint8_t ff_dither_8x8_32[9][8];
851 extern const uint8_t ff_dither_8x8_73[9][8];
852 extern const uint8_t ff_dither_8x8_128[9][8];
853 extern const uint8_t ff_dither_8x8_220[9][8];
855 extern const int32_t ff_yuv2rgb_coeffs[11][4];
857 extern const AVClass ff_sws_context_class;
860 * Set c->swscale to an unscaled converter if one exists for the specific
861 * source and destination formats, bit depths, flags, etc.
863 void ff_get_unscaled_swscale(SwsContext *c);
864 void ff_get_unscaled_swscale_ppc(SwsContext *c);
865 void ff_get_unscaled_swscale_arm(SwsContext *c);
866 void ff_get_unscaled_swscale_aarch64(SwsContext *c);
869 * Return function pointer to fastest main scaler path function depending
870 * on architecture and available optimizations.
872 SwsFunc ff_getSwsFunc(SwsContext *c);
874 void ff_sws_init_input_funcs(SwsContext *c);
875 void ff_sws_init_output_funcs(SwsContext *c,
876 yuv2planar1_fn *yuv2plane1,
877 yuv2planarX_fn *yuv2planeX,
878 yuv2interleavedX_fn *yuv2nv12cX,
879 yuv2packed1_fn *yuv2packed1,
880 yuv2packed2_fn *yuv2packed2,
881 yuv2packedX_fn *yuv2packedX,
882 yuv2anyX_fn *yuv2anyX);
883 void ff_sws_init_swscale_ppc(SwsContext *c);
884 void ff_sws_init_swscale_vsx(SwsContext *c);
885 void ff_sws_init_swscale_x86(SwsContext *c);
886 void ff_sws_init_swscale_aarch64(SwsContext *c);
887 void ff_sws_init_swscale_arm(SwsContext *c);
889 void ff_hyscale_fast_c(SwsContext *c, int16_t *dst, int dstWidth,
890 const uint8_t *src, int srcW, int xInc);
891 void ff_hcscale_fast_c(SwsContext *c, int16_t *dst1, int16_t *dst2,
892 int dstWidth, const uint8_t *src1,
893 const uint8_t *src2, int srcW, int xInc);
894 int ff_init_hscaler_mmxext(int dstW, int xInc, uint8_t *filterCode,
895 int16_t *filter, int32_t *filterPos,
897 void ff_hyscale_fast_mmxext(SwsContext *c, int16_t *dst,
898 int dstWidth, const uint8_t *src,
900 void ff_hcscale_fast_mmxext(SwsContext *c, int16_t *dst1, int16_t *dst2,
901 int dstWidth, const uint8_t *src1,
902 const uint8_t *src2, int srcW, int xInc);
905 * Allocate and return an SwsContext.
906 * This is like sws_getContext() but does not perform the init step, allowing
907 * the user to set additional AVOptions.
909 * @see sws_getContext()
911 struct SwsContext *sws_alloc_set_opts(int srcW, int srcH, enum AVPixelFormat srcFormat,
912 int dstW, int dstH, enum AVPixelFormat dstFormat,
913 int flags, const double *param);
915 int ff_sws_alphablendaway(SwsContext *c, const uint8_t *src[],
916 int srcStride[], int srcSliceY, int srcSliceH,
917 uint8_t *dst[], int dstStride[]);
919 static inline void fillPlane16(uint8_t *plane, int stride, int width, int height, int y,
920 int alpha, int bits, const int big_endian)
923 uint8_t *ptr = plane + stride * y;
924 int v = alpha ? 0xFFFF>>(16-bits) : (1<<(bits-1));
925 for (i = 0; i < height; i++) {
926 #define FILL(wfunc) \
927 for (j = 0; j < width; j++) {\
940 static inline void fillPlane32(uint8_t *plane, int stride, int width, int height, int y,
941 int alpha, int bits, const int big_endian, int is_float)
944 uint8_t *ptr = plane + stride * y;
946 uint32_t onef32 = 0x3f800000;
948 v = alpha ? onef32 : 0;
950 v = alpha ? 0xFFFFFFFF>>(32-bits) : (1<<(bits-1));
952 for (i = 0; i < height; i++) {
953 #define FILL(wfunc) \
954 for (j = 0; j < width; j++) {\
968 #define MAX_SLICE_PLANES 4
971 typedef struct SwsPlane
973 int available_lines; ///< max number of lines that can be hold by this plane
974 int sliceY; ///< index of first line
975 int sliceH; ///< number of lines
976 uint8_t **line; ///< line buffer
977 uint8_t **tmp; ///< Tmp line buffer used by mmx code
981 * Struct which defines a slice of an image to be scaled or an output for
983 * A slice can also be used as intermediate ring buffer for scaling steps.
985 typedef struct SwsSlice
987 int width; ///< Slice line width
988 int h_chr_sub_sample; ///< horizontal chroma subsampling factor
989 int v_chr_sub_sample; ///< vertical chroma subsampling factor
990 int is_ring; ///< flag to identify if this slice is a ring buffer
991 int should_free_lines; ///< flag to identify if there are dynamic allocated lines
992 enum AVPixelFormat fmt; ///< planes pixel format
993 SwsPlane plane[MAX_SLICE_PLANES]; ///< color planes
997 * Struct which holds all necessary data for processing a slice.
998 * A processing step can be a color conversion or horizontal/vertical scaling.
1000 typedef struct SwsFilterDescriptor
1002 SwsSlice *src; ///< Source slice
1003 SwsSlice *dst; ///< Output slice
1005 int alpha; ///< Flag for processing alpha channel
1006 void *instance; ///< Filter instance data
1008 /// Function for processing input slice sliceH lines starting from line sliceY
1009 int (*process)(SwsContext *c, struct SwsFilterDescriptor *desc, int sliceY, int sliceH);
1010 } SwsFilterDescriptor;
1012 // warp input lines in the form (src + width*i + j) to slice format (line[i][j])
1013 // relative=true means first line src[x][0] otherwise first line is src[x][lum/crh Y]
1014 int ff_init_slice_from_src(SwsSlice * s, uint8_t *src[4], int stride[4], int srcW, int lumY, int lumH, int chrY, int chrH, int relative);
1016 // Initialize scaler filter descriptor chain
1017 int ff_init_filters(SwsContext *c);
1019 // Free all filter data
1020 int ff_free_filters(SwsContext *c);
1023 function for applying ring buffer logic into slice s
1024 It checks if the slice can hold more @lum lines, if yes
1025 do nothing otherwise remove @lum least used lines.
1026 It applies the same procedure for @chr lines.
1028 int ff_rotate_slice(SwsSlice *s, int lum, int chr);
1030 /// initializes gamma conversion descriptor
1031 int ff_init_gamma_convert(SwsFilterDescriptor *desc, SwsSlice * src, uint16_t *table);
1033 /// initializes lum pixel format conversion descriptor
1034 int ff_init_desc_fmt_convert(SwsFilterDescriptor *desc, SwsSlice * src, SwsSlice *dst, uint32_t *pal);
1036 /// initializes lum horizontal scaling descriptor
1037 int ff_init_desc_hscale(SwsFilterDescriptor *desc, SwsSlice *src, SwsSlice *dst, uint16_t *filter, int * filter_pos, int filter_size, int xInc);
1039 /// initializes chr pixel format conversion descriptor
1040 int ff_init_desc_cfmt_convert(SwsFilterDescriptor *desc, SwsSlice * src, SwsSlice *dst, uint32_t *pal);
1042 /// initializes chr horizontal scaling descriptor
1043 int ff_init_desc_chscale(SwsFilterDescriptor *desc, SwsSlice *src, SwsSlice *dst, uint16_t *filter, int * filter_pos, int filter_size, int xInc);
1045 int ff_init_desc_no_chr(SwsFilterDescriptor *desc, SwsSlice * src, SwsSlice *dst);
1047 /// initializes vertical scaling descriptors
1048 int ff_init_vscale(SwsContext *c, SwsFilterDescriptor *desc, SwsSlice *src, SwsSlice *dst);
1050 /// setup vertical scaler functions
1051 void ff_init_vscale_pfn(SwsContext *c, yuv2planar1_fn yuv2plane1, yuv2planarX_fn yuv2planeX,
1052 yuv2interleavedX_fn yuv2nv12cX, yuv2packed1_fn yuv2packed1, yuv2packed2_fn yuv2packed2,
1053 yuv2packedX_fn yuv2packedX, yuv2anyX_fn yuv2anyX, int use_mmx);
1055 //number of extra lines to process
1056 #define MAX_LINES_AHEAD 4
1058 #endif /* SWSCALE_SWSCALE_INTERNAL_H */