falkTX
/
FFmpeg
mirror of https://github.com/falkTX/FFmpeg.git
1
0
Fork 0
Code Issues 0 Releases 338 Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
100996 Commits
32 Branches
824MB
Tree: 3044d0efee
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '3044d0efee'
${ noResults }
FFmpeg/libavfilter/drawutils.c

646 lines
24KB
Raw Blame History 

  1. /*

  2.  * Copyright 2011 Stefano Sabatini <stefano.sabatini-lala poste it>

  3.  * Copyright 2012 Nicolas George <nicolas.george normalesup org>

  4.  *

  5.  * This file is part of FFmpeg.

  6.  *

  7.  * FFmpeg is free software; you can redistribute it and/or

  8.  * modify it under the terms of the GNU Lesser General Public

  9.  * License as published by the Free Software Foundation; either

  10.  * version 2.1 of the License, or (at your option) any later version.

  11.  *

  12.  * FFmpeg is distributed in the hope that it will be useful,

  13.  * but WITHOUT ANY WARRANTY; without even the implied warranty of

  14.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU

  15.  * Lesser General Public License for more details.

  16.  *

  17.  * You should have received a copy of the GNU Lesser General Public

  18.  * License along with FFmpeg; if not, write to the Free Software

  19.  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

  20.  */

  21. 

  22. #include <string.h>

  23. 

  24. #include "libavutil/avassert.h"

  25. #include "libavutil/avutil.h"

  26. #include "libavutil/colorspace.h"

  27. #include "libavutil/intreadwrite.h"

  28. #include "libavutil/mem.h"

  29. #include "libavutil/pixdesc.h"

  30. #include "drawutils.h"

  31. #include "formats.h"

  32. 

  33. enum { RED = 0, GREEN, BLUE, ALPHA };

  34. 

  35. int ff_fill_rgba_map(uint8_t *rgba_map, enum AVPixelFormat pix_fmt)

  36. {

  37.     switch (pix_fmt) {

  38.     case AV_PIX_FMT_0RGB:

  39.     case AV_PIX_FMT_ARGB:  rgba_map[ALPHA] = 0; rgba_map[RED  ] = 1; rgba_map[GREEN] = 2; rgba_map[BLUE ] = 3; break;

  40.     case AV_PIX_FMT_0BGR:

  41.     case AV_PIX_FMT_ABGR:  rgba_map[ALPHA] = 0; rgba_map[BLUE ] = 1; rgba_map[GREEN] = 2; rgba_map[RED  ] = 3; break;

  42.     case AV_PIX_FMT_RGB48LE:

  43.     case AV_PIX_FMT_RGB48BE:

  44.     case AV_PIX_FMT_RGBA64BE:

  45.     case AV_PIX_FMT_RGBA64LE:

  46.     case AV_PIX_FMT_RGB0:

  47.     case AV_PIX_FMT_RGBA:

  48.     case AV_PIX_FMT_RGB24: rgba_map[RED  ] = 0; rgba_map[GREEN] = 1; rgba_map[BLUE ] = 2; rgba_map[ALPHA] = 3; break;

  49.     case AV_PIX_FMT_BGR48LE:

  50.     case AV_PIX_FMT_BGR48BE:

  51.     case AV_PIX_FMT_BGRA64BE:

  52.     case AV_PIX_FMT_BGRA64LE:

  53.     case AV_PIX_FMT_BGRA:

  54.     case AV_PIX_FMT_BGR0:

  55.     case AV_PIX_FMT_BGR24: rgba_map[BLUE ] = 0; rgba_map[GREEN] = 1; rgba_map[RED  ] = 2; rgba_map[ALPHA] = 3; break;

  56.     case AV_PIX_FMT_GBRP9LE:

  57.     case AV_PIX_FMT_GBRP9BE:

  58.     case AV_PIX_FMT_GBRP10LE:

  59.     case AV_PIX_FMT_GBRP10BE:

  60.     case AV_PIX_FMT_GBRP12LE:

  61.     case AV_PIX_FMT_GBRP12BE:

  62.     case AV_PIX_FMT_GBRP14LE:

  63.     case AV_PIX_FMT_GBRP14BE:

  64.     case AV_PIX_FMT_GBRP16LE:

  65.     case AV_PIX_FMT_GBRP16BE:

  66.     case AV_PIX_FMT_GBRAP:

  67.     case AV_PIX_FMT_GBRAP10LE:

  68.     case AV_PIX_FMT_GBRAP10BE:

  69.     case AV_PIX_FMT_GBRAP12LE:

  70.     case AV_PIX_FMT_GBRAP12BE:

  71.     case AV_PIX_FMT_GBRAP16LE:

  72.     case AV_PIX_FMT_GBRAP16BE:

  73.     case AV_PIX_FMT_GBRP:  rgba_map[GREEN] = 0; rgba_map[BLUE ] = 1; rgba_map[RED  ] = 2; rgba_map[ALPHA] = 3; break;

  74.     default:                    /* unsupported */

  75.         return AVERROR(EINVAL);

  76.     }

  77.     return 0;

  78. }

  79. 

  80. int ff_draw_init(FFDrawContext *draw, enum AVPixelFormat format, unsigned flags)

  81. {

  82.     const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(format);

  83.     const AVComponentDescriptor *c;

  84.     unsigned i, nb_planes = 0;

  85.     int pixelstep[MAX_PLANES] = { 0 };

  86.     int full_range = 0;

  87. 

  88.     if (!desc || !desc->name)

  89.         return AVERROR(EINVAL);

  90.     if (desc->flags & ~(AV_PIX_FMT_FLAG_PLANAR | AV_PIX_FMT_FLAG_RGB | FF_PSEUDOPAL | AV_PIX_FMT_FLAG_ALPHA))

  91.         return AVERROR(ENOSYS);

  92.     if (format == AV_PIX_FMT_P010LE || format == AV_PIX_FMT_P010BE || format == AV_PIX_FMT_P016LE || format == AV_PIX_FMT_P016BE)

  93.         return AVERROR(ENOSYS);

  94.     if (format == AV_PIX_FMT_YUVJ420P || format == AV_PIX_FMT_YUVJ422P || format == AV_PIX_FMT_YUVJ444P ||

  95.         format == AV_PIX_FMT_YUVJ411P || format == AV_PIX_FMT_YUVJ440P)

  96.         full_range = 1;

  97.     for (i = 0; i < desc->nb_components; i++) {

  98.         c = &desc->comp[i];

  99.         /* for now, only 8-16 bits formats */

  100.         if (c->depth < 8 || c->depth > 16)

  101.             return AVERROR(ENOSYS);

  102.         if (desc->flags & AV_PIX_FMT_FLAG_BE)

  103.             return AVERROR(ENOSYS);

  104.         if (c->plane >= MAX_PLANES)

  105.             return AVERROR(ENOSYS);

  106.         /* strange interleaving */

  107.         if (pixelstep[c->plane] != 0 &&

  108.             pixelstep[c->plane] != c->step)

  109.             return AVERROR(ENOSYS);

  110.         if (pixelstep[c->plane] == 6 &&

  111.             c->depth == 16)

  112.             return AVERROR(ENOSYS);

  113.         pixelstep[c->plane] = c->step;

  114.         if (pixelstep[c->plane] >= 8)

  115.             return AVERROR(ENOSYS);

  116.         nb_planes = FFMAX(nb_planes, c->plane + 1);

  117.     }

  118.     memset(draw, 0, sizeof(*draw));

  119.     draw->desc      = desc;

  120.     draw->format    = format;

  121.     draw->nb_planes = nb_planes;

  122.     draw->flags     = flags;

  123.     draw->full_range = full_range;

  124.     memcpy(draw->pixelstep, pixelstep, sizeof(draw->pixelstep));

  125.     draw->hsub[1] = draw->hsub[2] = draw->hsub_max = desc->log2_chroma_w;

  126.     draw->vsub[1] = draw->vsub[2] = draw->vsub_max = desc->log2_chroma_h;

  127.     for (i = 0; i < (desc->nb_components - !!(desc->flags & AV_PIX_FMT_FLAG_ALPHA && !(flags & FF_DRAW_PROCESS_ALPHA))); i++)

  128.         draw->comp_mask[desc->comp[i].plane] |=

  129.             1 << desc->comp[i].offset;

  130.     return 0;

  131. }

  132. 

  133. void ff_draw_color(FFDrawContext *draw, FFDrawColor *color, const uint8_t rgba[4])

  134. {

  135.     unsigned i;

  136.     uint8_t rgba_map[4];

  137. 

  138.     if (rgba != color->rgba)

  139.         memcpy(color->rgba, rgba, sizeof(color->rgba));

  140.     if ((draw->desc->flags & AV_PIX_FMT_FLAG_RGB) &&

  141.         ff_fill_rgba_map(rgba_map, draw->format) >= 0) {

  142.         if (draw->nb_planes == 1) {

  143.             for (i = 0; i < 4; i++) {

  144.                 color->comp[0].u8[rgba_map[i]] = rgba[i];

  145.                 if (draw->desc->comp[rgba_map[i]].depth > 8) {

  146.                     color->comp[0].u16[rgba_map[i]] = color->comp[0].u8[rgba_map[i]] << 8;

  147.                 }

  148.             }

  149.         } else {

  150.             for (i = 0; i < 4; i++) {

  151.                 color->comp[rgba_map[i]].u8[0] = rgba[i];

  152.                 if (draw->desc->comp[rgba_map[i]].depth > 8)

  153.                     color->comp[rgba_map[i]].u16[0] = color->comp[rgba_map[i]].u8[0] << (draw->desc->comp[rgba_map[i]].depth - 8);

  154.             }

  155.         }

  156.     } else if (draw->nb_planes >= 2) {

  157.         /* assume YUV */

  158.         const AVPixFmtDescriptor *desc = draw->desc;

  159.         color->comp[desc->comp[0].plane].u8[desc->comp[0].offset] = draw->full_range ? RGB_TO_Y_JPEG(rgba[0], rgba[1], rgba[2]) : RGB_TO_Y_CCIR(rgba[0], rgba[1], rgba[2]);

  160.         color->comp[desc->comp[1].plane].u8[desc->comp[1].offset] = draw->full_range ? RGB_TO_U_JPEG(rgba[0], rgba[1], rgba[2]) : RGB_TO_U_CCIR(rgba[0], rgba[1], rgba[2], 0);

  161.         color->comp[desc->comp[2].plane].u8[desc->comp[2].offset] = draw->full_range ? RGB_TO_V_JPEG(rgba[0], rgba[1], rgba[2]) : RGB_TO_V_CCIR(rgba[0], rgba[1], rgba[2], 0);

  162.         color->comp[3].u8[0] = rgba[3];

  163. #define EXPAND(compn) \

  164.         if (desc->comp[compn].depth > 8) \

  165.             color->comp[desc->comp[compn].plane].u16[desc->comp[compn].offset] = \

  166.             color->comp[desc->comp[compn].plane].u8[desc->comp[compn].offset] << \

  167.                 (draw->desc->comp[compn].depth + draw->desc->comp[compn].shift - 8)

  168.         EXPAND(3);

  169.         EXPAND(2);

  170.         EXPAND(1);

  171.         EXPAND(0);

  172.     } else if (draw->format == AV_PIX_FMT_GRAY8 || draw->format == AV_PIX_FMT_GRAY8A ||

  173.                draw->format == AV_PIX_FMT_GRAY16LE || draw->format == AV_PIX_FMT_YA16LE ||

  174.                draw->format == AV_PIX_FMT_GRAY9LE  ||

  175.                draw->format == AV_PIX_FMT_GRAY10LE ||

  176.                draw->format == AV_PIX_FMT_GRAY12LE ||

  177.                draw->format == AV_PIX_FMT_GRAY14LE) {

  178.         const AVPixFmtDescriptor *desc = draw->desc;

  179.         color->comp[0].u8[0] = RGB_TO_Y_CCIR(rgba[0], rgba[1], rgba[2]);

  180.         EXPAND(0);

  181.         color->comp[1].u8[0] = rgba[3];

  182.         EXPAND(1);

  183.     } else {

  184.         av_log(NULL, AV_LOG_WARNING,

  185.                "Color conversion not implemented for %s\n", draw->desc->name);

  186.         memset(color, 128, sizeof(*color));

  187.     }

  188. }

  189. 

  190. static uint8_t *pointer_at(FFDrawContext *draw, uint8_t *data[], int linesize[],

  191.                            int plane, int x, int y)

  192. {

  193.     return data[plane] +

  194.            (y >> draw->vsub[plane]) * linesize[plane] +

  195.            (x >> draw->hsub[plane]) * draw->pixelstep[plane];

  196. }

  197. 

  198. void ff_copy_rectangle2(FFDrawContext *draw,

  199.                         uint8_t *dst[], int dst_linesize[],

  200.                         uint8_t *src[], int src_linesize[],

  201.                         int dst_x, int dst_y, int src_x, int src_y,

  202.                         int w, int h)

  203. {

  204.     int plane, y, wp, hp;

  205.     uint8_t *p, *q;

  206. 

  207.     for (plane = 0; plane < draw->nb_planes; plane++) {

  208.         p = pointer_at(draw, src, src_linesize, plane, src_x, src_y);

  209.         q = pointer_at(draw, dst, dst_linesize, plane, dst_x, dst_y);

  210.         wp = AV_CEIL_RSHIFT(w, draw->hsub[plane]) * draw->pixelstep[plane];

  211.         hp = AV_CEIL_RSHIFT(h, draw->vsub[plane]);

  212.         for (y = 0; y < hp; y++) {

  213.             memcpy(q, p, wp);

  214.             p += src_linesize[plane];

  215.             q += dst_linesize[plane];

  216.         }

  217.     }

  218. }

  219. 

  220. void ff_fill_rectangle(FFDrawContext *draw, FFDrawColor *color,

  221.                        uint8_t *dst[], int dst_linesize[],

  222.                        int dst_x, int dst_y, int w, int h)

  223. {

  224.     int plane, x, y, wp, hp;

  225.     uint8_t *p0, *p;

  226.     FFDrawColor color_tmp = *color;

  227. 

  228.     for (plane = 0; plane < draw->nb_planes; plane++) {

  229.         p0 = pointer_at(draw, dst, dst_linesize, plane, dst_x, dst_y);

  230.         wp = AV_CEIL_RSHIFT(w, draw->hsub[plane]);

  231.         hp = AV_CEIL_RSHIFT(h, draw->vsub[plane]);

  232.         if (!hp)

  233.             return;

  234.         p = p0;

  235. 

  236.         if (HAVE_BIGENDIAN && draw->desc->comp[0].depth > 8) {

  237.             for (x = 0; 2*x < draw->pixelstep[plane]; x++)

  238.                 color_tmp.comp[plane].u16[x] = av_bswap16(color_tmp.comp[plane].u16[x]);

  239.         }

  240. 

  241.         /* copy first line from color */

  242.         for (x = 0; x < wp; x++) {

  243.             memcpy(p, color_tmp.comp[plane].u8, draw->pixelstep[plane]);

  244.             p += draw->pixelstep[plane];

  245.         }

  246.         wp *= draw->pixelstep[plane];

  247.         /* copy next lines from first line */

  248.         p = p0 + dst_linesize[plane];

  249.         for (y = 1; y < hp; y++) {

  250.             memcpy(p, p0, wp);

  251.             p += dst_linesize[plane];

  252.         }

  253.     }

  254. }

  255. 

  256. /**

  257.  * Clip interval [x; x+w[ within [0; wmax[.

  258.  * The resulting w may be negative if the final interval is empty.

  259.  * dx, if not null, return the difference between in and out value of x.

  260.  */

  261. static void clip_interval(int wmax, int *x, int *w, int *dx)

  262. {

  263.     if (dx)

  264.         *dx = 0;

  265.     if (*x < 0) {

  266.         if (dx)

  267.             *dx = -*x;

  268.         *w += *x;

  269.         *x = 0;

  270.     }

  271.     if (*x + *w > wmax)

  272.         *w = wmax - *x;

  273. }

  274. 

  275. /**

  276.  * Decompose w pixels starting at x

  277.  * into start + (w starting at x) + end

  278.  * with x and w aligned on multiples of 1<<sub.

  279.  */

  280. static void subsampling_bounds(int sub, int *x, int *w, int *start, int *end)

  281. {

  282.     int mask = (1 << sub) - 1;

  283. 

  284.     *start = (-*x) & mask;

  285.     *x += *start;

  286.     *start = FFMIN(*start, *w);

  287.     *w -= *start;

  288.     *end = *w & mask;

  289.     *w >>= sub;

  290. }

  291. 

  292. static int component_used(FFDrawContext *draw, int plane, int comp)

  293. {

  294.     return (draw->comp_mask[plane] >> comp) & 1;

  295. }

  296. 

  297. /* If alpha is in the [ 0 ; 0x1010101 ] range,

  298.    then alpha * value is in the [ 0 ; 0xFFFFFFFF ] range,

  299.    and >> 24 gives a correct rounding. */

  300. static void blend_line(uint8_t *dst, unsigned src, unsigned alpha,

  301.                        int dx, int w, unsigned hsub, int left, int right)

  302. {

  303.     unsigned asrc = alpha * src;

  304.     unsigned tau = 0x1010101 - alpha;

  305.     int x;

  306. 

  307.     if (left) {

  308.         unsigned suba = (left * alpha) >> hsub;

  309.         *dst = (*dst * (0x1010101 - suba) + src * suba) >> 24;

  310.         dst += dx;

  311.     }

  312.     for (x = 0; x < w; x++) {

  313.         *dst = (*dst * tau + asrc) >> 24;

  314.         dst += dx;

  315.     }

  316.     if (right) {

  317.         unsigned suba = (right * alpha) >> hsub;

  318.         *dst = (*dst * (0x1010101 - suba) + src * suba) >> 24;

  319.     }

  320. }

  321. 

  322. static void blend_line16(uint8_t *dst, unsigned src, unsigned alpha,

  323.                          int dx, int w, unsigned hsub, int left, int right)

  324. {

  325.     unsigned asrc = alpha * src;

  326.     unsigned tau = 0x10001 - alpha;

  327.     int x;

  328. 

  329.     if (left) {

  330.         unsigned suba = (left * alpha) >> hsub;

  331.         uint16_t value = AV_RL16(dst);

  332.         AV_WL16(dst, (value * (0x10001 - suba) + src * suba) >> 16);

  333.         dst += dx;

  334.     }

  335.     for (x = 0; x < w; x++) {

  336.         uint16_t value = AV_RL16(dst);

  337.         AV_WL16(dst, (value * tau + asrc) >> 16);

  338.         dst += dx;

  339.     }

  340.     if (right) {

  341.         unsigned suba = (right * alpha) >> hsub;

  342.         uint16_t value = AV_RL16(dst);

  343.         AV_WL16(dst, (value * (0x10001 - suba) + src * suba) >> 16);

  344.     }

  345. }

  346. 

  347. void ff_blend_rectangle(FFDrawContext *draw, FFDrawColor *color,

  348.                         uint8_t *dst[], int dst_linesize[],

  349.                         int dst_w, int dst_h,

  350.                         int x0, int y0, int w, int h)

  351. {

  352.     unsigned alpha, nb_planes, nb_comp, plane, comp;

  353.     int w_sub, h_sub, x_sub, y_sub, left, right, top, bottom, y;

  354.     uint8_t *p0, *p;

  355. 

  356.     /* TODO optimize if alpha = 0xFF */

  357.     clip_interval(dst_w, &x0, &w, NULL);

  358.     clip_interval(dst_h, &y0, &h, NULL);

  359.     if (w <= 0 || h <= 0 || !color->rgba[3])

  360.         return;

  361.     if (draw->desc->comp[0].depth <= 8) {

  362.         /* 0x10203 * alpha + 2 is in the [ 2 ; 0x1010101 - 2 ] range */

  363.         alpha = 0x10203 * color->rgba[3] + 0x2;

  364.     } else {

  365.         /* 0x101 * alpha is in the [ 2 ; 0x1001] range */

  366.         alpha = 0x101 * color->rgba[3] + 0x2;

  367.     }

  368.     nb_planes = draw->nb_planes - !!(draw->desc->flags & AV_PIX_FMT_FLAG_ALPHA && !(draw->flags & FF_DRAW_PROCESS_ALPHA));

  369.     nb_planes += !nb_planes;

  370.     for (plane = 0; plane < nb_planes; plane++) {

  371.         nb_comp = draw->pixelstep[plane];

  372.         p0 = pointer_at(draw, dst, dst_linesize, plane, x0, y0);

  373.         w_sub = w;

  374.         h_sub = h;

  375.         x_sub = x0;

  376.         y_sub = y0;

  377.         subsampling_bounds(draw->hsub[plane], &x_sub, &w_sub, &left, &right);

  378.         subsampling_bounds(draw->vsub[plane], &y_sub, &h_sub, &top, &bottom);

  379.         for (comp = 0; comp < nb_comp; comp++) {

  380.             const int depth = draw->desc->comp[comp].depth;

  381. 

  382.             if (!component_used(draw, plane, comp))

  383.                 continue;

  384.             p = p0 + comp;

  385.             if (top) {

  386.                 if (depth <= 8) {

  387.                     blend_line(p, color->comp[plane].u8[comp], alpha >> 1,

  388.                                draw->pixelstep[plane], w_sub,

  389.                                draw->hsub[plane], left, right);

  390.                 } else {

  391.                     blend_line16(p, color->comp[plane].u16[comp], alpha >> 1,

  392.                                  draw->pixelstep[plane], w_sub,

  393.                                  draw->hsub[plane], left, right);

  394.                 }

  395.                 p += dst_linesize[plane];

  396.             }

  397.             if (depth <= 8) {

  398.                 for (y = 0; y < h_sub; y++) {

  399.                     blend_line(p, color->comp[plane].u8[comp], alpha,

  400.                                draw->pixelstep[plane], w_sub,

  401.                                draw->hsub[plane], left, right);

  402.                     p += dst_linesize[plane];

  403.                 }

  404.             } else {

  405.                 for (y = 0; y < h_sub; y++) {

  406.                     blend_line16(p, color->comp[plane].u16[comp], alpha,

  407.                                  draw->pixelstep[plane], w_sub,

  408.                                  draw->hsub[plane], left, right);

  409.                     p += dst_linesize[plane];

  410.                 }

  411.             }

  412.             if (bottom) {

  413.                 if (depth <= 8) {

  414.                     blend_line(p, color->comp[plane].u8[comp], alpha >> 1,

  415.                                draw->pixelstep[plane], w_sub,

  416.                                draw->hsub[plane], left, right);

  417.                 } else {

  418.                     blend_line16(p, color->comp[plane].u16[comp], alpha >> 1,

  419.                                  draw->pixelstep[plane], w_sub,

  420.                                  draw->hsub[plane], left, right);

  421.                 }

  422.             }

  423.         }

  424.     }

  425. }

  426. 

  427. static void blend_pixel16(uint8_t *dst, unsigned src, unsigned alpha,

  428.                           const uint8_t *mask, int mask_linesize, int l2depth,

  429.                           unsigned w, unsigned h, unsigned shift, unsigned xm0)

  430. {

  431.     unsigned xm, x, y, t = 0;

  432.     unsigned xmshf = 3 - l2depth;

  433.     unsigned xmmod = 7 >> l2depth;

  434.     unsigned mbits = (1 << (1 << l2depth)) - 1;

  435.     unsigned mmult = 255 / mbits;

  436.     uint16_t value = AV_RL16(dst);

  437. 

  438.     for (y = 0; y < h; y++) {

  439.         xm = xm0;

  440.         for (x = 0; x < w; x++) {

  441.             t += ((mask[xm >> xmshf] >> ((~xm & xmmod) << l2depth)) & mbits)

  442.                  * mmult;

  443.             xm++;

  444.         }

  445.         mask += mask_linesize;

  446.     }

  447.     alpha = (t >> shift) * alpha;

  448.     AV_WL16(dst, ((0x10001 - alpha) * value + alpha * src) >> 16);

  449. }

  450. 

  451. static void blend_pixel(uint8_t *dst, unsigned src, unsigned alpha,

  452.                         const uint8_t *mask, int mask_linesize, int l2depth,

  453.                         unsigned w, unsigned h, unsigned shift, unsigned xm0)

  454. {

  455.     unsigned xm, x, y, t = 0;

  456.     unsigned xmshf = 3 - l2depth;

  457.     unsigned xmmod = 7 >> l2depth;

  458.     unsigned mbits = (1 << (1 << l2depth)) - 1;

  459.     unsigned mmult = 255 / mbits;

  460. 

  461.     for (y = 0; y < h; y++) {

  462.         xm = xm0;

  463.         for (x = 0; x < w; x++) {

  464.             t += ((mask[xm >> xmshf] >> ((~xm & xmmod) << l2depth)) & mbits)

  465.                  * mmult;

  466.             xm++;

  467.         }

  468.         mask += mask_linesize;

  469.     }

  470.     alpha = (t >> shift) * alpha;

  471.     *dst = ((0x1010101 - alpha) * *dst + alpha * src) >> 24;

  472. }

  473. 

  474. static void blend_line_hv16(uint8_t *dst, int dst_delta,

  475.                             unsigned src, unsigned alpha,

  476.                             const uint8_t *mask, int mask_linesize, int l2depth, int w,

  477.                             unsigned hsub, unsigned vsub,

  478.                             int xm, int left, int right, int hband)

  479. {

  480.     int x;

  481. 

  482.     if (left) {

  483.         blend_pixel16(dst, src, alpha, mask, mask_linesize, l2depth,

  484.                       left, hband, hsub + vsub, xm);

  485.         dst += dst_delta;

  486.         xm += left;

  487.     }

  488.     for (x = 0; x < w; x++) {

  489.         blend_pixel16(dst, src, alpha, mask, mask_linesize, l2depth,

  490.                       1 << hsub, hband, hsub + vsub, xm);

  491.         dst += dst_delta;

  492.         xm += 1 << hsub;

  493.     }

  494.     if (right)

  495.         blend_pixel16(dst, src, alpha, mask, mask_linesize, l2depth,

  496.                       right, hband, hsub + vsub, xm);

  497. }

  498. 

  499. static void blend_line_hv(uint8_t *dst, int dst_delta,

  500.                           unsigned src, unsigned alpha,

  501.                           const uint8_t *mask, int mask_linesize, int l2depth, int w,

  502.                           unsigned hsub, unsigned vsub,

  503.                           int xm, int left, int right, int hband)

  504. {

  505.     int x;

  506. 

  507.     if (left) {

  508.         blend_pixel(dst, src, alpha, mask, mask_linesize, l2depth,

  509.                     left, hband, hsub + vsub, xm);

  510.         dst += dst_delta;

  511.         xm += left;

  512.     }

  513.     for (x = 0; x < w; x++) {

  514.         blend_pixel(dst, src, alpha, mask, mask_linesize, l2depth,

  515.                     1 << hsub, hband, hsub + vsub, xm);

  516.         dst += dst_delta;

  517.         xm += 1 << hsub;

  518.     }

  519.     if (right)

  520.         blend_pixel(dst, src, alpha, mask, mask_linesize, l2depth,

  521.                     right, hband, hsub + vsub, xm);

  522. }

  523. 

  524. void ff_blend_mask(FFDrawContext *draw, FFDrawColor *color,

  525.                    uint8_t *dst[], int dst_linesize[], int dst_w, int dst_h,

  526.                    const uint8_t *mask,  int mask_linesize, int mask_w, int mask_h,

  527.                    int l2depth, unsigned endianness, int x0, int y0)

  528. {

  529.     unsigned alpha, nb_planes, nb_comp, plane, comp;

  530.     int xm0, ym0, w_sub, h_sub, x_sub, y_sub, left, right, top, bottom, y;

  531.     uint8_t *p0, *p;

  532.     const uint8_t *m;

  533. 

  534.     clip_interval(dst_w, &x0, &mask_w, &xm0);

  535.     clip_interval(dst_h, &y0, &mask_h, &ym0);

  536.     mask += ym0 * mask_linesize;

  537.     if (mask_w <= 0 || mask_h <= 0 || !color->rgba[3])

  538.         return;

  539.     if (draw->desc->comp[0].depth <= 8) {

  540.         /* alpha is in the [ 0 ; 0x10203 ] range,

  541.            alpha * mask is in the [ 0 ; 0x1010101 - 4 ] range */

  542.         alpha = (0x10307 * color->rgba[3] + 0x3) >> 8;

  543.     } else {

  544.         alpha = (0x101 * color->rgba[3] + 0x2) >> 8;

  545.     }

  546.     nb_planes = draw->nb_planes - !!(draw->desc->flags & AV_PIX_FMT_FLAG_ALPHA && !(draw->flags & FF_DRAW_PROCESS_ALPHA));

  547.     nb_planes += !nb_planes;

  548.     for (plane = 0; plane < nb_planes; plane++) {

  549.         nb_comp = draw->pixelstep[plane];

  550.         p0 = pointer_at(draw, dst, dst_linesize, plane, x0, y0);

  551.         w_sub = mask_w;

  552.         h_sub = mask_h;

  553.         x_sub = x0;

  554.         y_sub = y0;

  555.         subsampling_bounds(draw->hsub[plane], &x_sub, &w_sub, &left, &right);

  556.         subsampling_bounds(draw->vsub[plane], &y_sub, &h_sub, &top, &bottom);

  557.         for (comp = 0; comp < nb_comp; comp++) {

  558.             const int depth = draw->desc->comp[comp].depth;

  559. 

  560.             if (!component_used(draw, plane, comp))

  561.                 continue;

  562.             p = p0 + comp;

  563.             m = mask;

  564.             if (top) {

  565.                 if (depth <= 8) {

  566.                     blend_line_hv(p, draw->pixelstep[plane],

  567.                                   color->comp[plane].u8[comp], alpha,

  568.                                   m, mask_linesize, l2depth, w_sub,

  569.                                   draw->hsub[plane], draw->vsub[plane],

  570.                                   xm0, left, right, top);

  571.                 } else {

  572.                     blend_line_hv16(p, draw->pixelstep[plane],

  573.                                     color->comp[plane].u16[comp], alpha,

  574.                                     m, mask_linesize, l2depth, w_sub,

  575.                                     draw->hsub[plane], draw->vsub[plane],

  576.                                     xm0, left, right, top);

  577.                 }

  578.                 p += dst_linesize[plane];

  579.                 m += top * mask_linesize;

  580.             }

  581.             if (depth <= 8) {

  582.                 for (y = 0; y < h_sub; y++) {

  583.                     blend_line_hv(p, draw->pixelstep[plane],

  584.                                   color->comp[plane].u8[comp], alpha,

  585.                                   m, mask_linesize, l2depth, w_sub,

  586.                                   draw->hsub[plane], draw->vsub[plane],

  587.                                   xm0, left, right, 1 << draw->vsub[plane]);

  588.                     p += dst_linesize[plane];

  589.                     m += mask_linesize << draw->vsub[plane];

  590.                 }

  591.             } else {

  592.                 for (y = 0; y < h_sub; y++) {

  593.                     blend_line_hv16(p, draw->pixelstep[plane],

  594.                                     color->comp[plane].u16[comp], alpha,

  595.                                     m, mask_linesize, l2depth, w_sub,

  596.                                     draw->hsub[plane], draw->vsub[plane],

  597.                                     xm0, left, right, 1 << draw->vsub[plane]);

  598.                     p += dst_linesize[plane];

  599.                     m += mask_linesize << draw->vsub[plane];

  600.                 }

  601.             }

  602.             if (bottom) {

  603.                 if (depth <= 8) {

  604.                     blend_line_hv(p, draw->pixelstep[plane],

  605.                                   color->comp[plane].u8[comp], alpha,

  606.                                   m, mask_linesize, l2depth, w_sub,

  607.                                   draw->hsub[plane], draw->vsub[plane],

  608.                                   xm0, left, right, bottom);

  609.                 } else {

  610.                     blend_line_hv16(p, draw->pixelstep[plane],

  611.                                     color->comp[plane].u16[comp], alpha,

  612.                                     m, mask_linesize, l2depth, w_sub,

  613.                                     draw->hsub[plane], draw->vsub[plane],

  614.                                     xm0, left, right, bottom);

  615.                 }

  616.             }

  617.         }

  618.     }

  619. }

  620. 

  621. int ff_draw_round_to_sub(FFDrawContext *draw, int sub_dir, int round_dir,

  622.                          int value)

  623. {

  624.     unsigned shift = sub_dir ? draw->vsub_max : draw->hsub_max;

  625. 

  626.     if (!shift)

  627.         return value;

  628.     if (round_dir >= 0)

  629.         value += round_dir ? (1 << shift) - 1 : 1 << (shift - 1);

  630.     return (value >> shift) << shift;

  631. }

  632. 

  633. AVFilterFormats *ff_draw_supported_pixel_formats(unsigned flags)

  634. {

  635.     enum AVPixelFormat i;

  636.     FFDrawContext draw;

  637.     AVFilterFormats *fmts = NULL;

  638.     int ret;

  639. 

  640.     for (i = 0; av_pix_fmt_desc_get(i); i++)

  641.         if (ff_draw_init(&draw, i, flags) >= 0 &&

  642.             (ret = ff_add_format(&fmts, i)) < 0)

  643.             return NULL;

  644.     return fmts;

  645. }