ffmpeg / libavcodec / flashsv2enc.c @ ba9c4db5
History | View | Annotate | Download (27.3 KB)
1 |
/*
|
---|---|
2 |
* Flash Screen Video Version 2 encoder
|
3 |
* Copyright (C) 2009 Joshua Warner
|
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 |
/**
|
23 |
* @file libavcodec/flashsv2enc.c
|
24 |
* Flash Screen Video Version 2 encoder
|
25 |
* @author Joshua Warner
|
26 |
*/
|
27 |
|
28 |
/* Differences from version 1 stream:
|
29 |
* NOTE: Currently, the only player that supports version 2 streams is Adobe Flash Player itself.
|
30 |
* * Supports sending only a range of scanlines in a block,
|
31 |
* indicating a difference from the corresponding block in the last keyframe.
|
32 |
* * Supports initializing the zlib dictionary with data from the corresponding
|
33 |
* block in the last keyframe, to improve compression.
|
34 |
* * Supports a hybrid 15-bit rgb / 7-bit palette color space.
|
35 |
*/
|
36 |
|
37 |
/* TODO:
|
38 |
* Don't keep Block structures for both current frame and keyframe.
|
39 |
* Make better heuristics for deciding stream parameters (optimum_* functions). Currently these return constants.
|
40 |
* Figure out how to encode palette information in the stream, choose an optimum palette at each keyframe.
|
41 |
* Figure out how the zlibPrimeCompressCurrent flag works, implement support.
|
42 |
* Find other sample files (that weren't generated here), develop a decoder.
|
43 |
*/
|
44 |
|
45 |
#include <stdio.h> |
46 |
#include <stdlib.h> |
47 |
#include <zlib.h> |
48 |
|
49 |
#include "libavutil/imgutils.h" |
50 |
#include "avcodec.h" |
51 |
#include "put_bits.h" |
52 |
#include "bytestream.h" |
53 |
|
54 |
#define HAS_IFRAME_IMAGE 0x02 |
55 |
#define HAS_PALLET_INFO 0x01 |
56 |
|
57 |
#define COLORSPACE_BGR 0x00 |
58 |
#define COLORSPACE_15_7 0x10 |
59 |
#define HAS_DIFF_BLOCKS 0x04 |
60 |
#define ZLIB_PRIME_COMPRESS_CURRENT 0x02 |
61 |
#define ZLIB_PRIME_COMPRESS_PREVIOUS 0x01 |
62 |
|
63 |
// Disables experimental "smart" parameter-choosing code, as well as the statistics that it depends on.
|
64 |
// At the moment, the "smart" code is a great example of how the parameters *shouldn't* be chosen.
|
65 |
#define FLASHSV2_DUMB
|
66 |
|
67 |
typedef struct Block { |
68 |
uint8_t *enc; |
69 |
uint8_t *sl_begin, *sl_end; |
70 |
int enc_size;
|
71 |
uint8_t *data; |
72 |
unsigned long data_size; |
73 |
|
74 |
uint8_t start, len; |
75 |
uint8_t dirty; |
76 |
uint8_t col, row, width, height; |
77 |
uint8_t flags; |
78 |
} Block; |
79 |
|
80 |
typedef struct Palette { |
81 |
unsigned colors[128]; |
82 |
uint8_t index[1 << 15]; |
83 |
} Palette; |
84 |
|
85 |
typedef struct FlashSV2Context { |
86 |
AVCodecContext *avctx; |
87 |
uint8_t *current_frame; |
88 |
uint8_t *key_frame; |
89 |
AVFrame frame; |
90 |
uint8_t *encbuffer; |
91 |
uint8_t *keybuffer; |
92 |
uint8_t *databuffer; |
93 |
|
94 |
Block *frame_blocks; |
95 |
Block *key_blocks; |
96 |
int frame_size;
|
97 |
int blocks_size;
|
98 |
|
99 |
int use15_7, dist, comp;
|
100 |
|
101 |
int rows, cols;
|
102 |
|
103 |
int last_key_frame;
|
104 |
|
105 |
int image_width, image_height;
|
106 |
int block_width, block_height;
|
107 |
uint8_t flags; |
108 |
uint8_t use_custom_palette; |
109 |
uint8_t palette_type; ///< 0=>default, 1=>custom - changed when palette regenerated.
|
110 |
Palette palette; |
111 |
#ifndef FLASHSV2_DUMB
|
112 |
double tot_blocks; ///< blocks encoded since last keyframe |
113 |
double diff_blocks; ///< blocks that were different since last keyframe |
114 |
double tot_lines; ///< total scanlines in image since last keyframe |
115 |
double diff_lines; ///< scanlines that were different since last keyframe |
116 |
double raw_size; ///< size of raw frames since last keyframe |
117 |
double comp_size; ///< size of compressed data since last keyframe |
118 |
double uncomp_size; ///< size of uncompressed data since last keyframe |
119 |
|
120 |
double total_bits; ///< total bits written to stream so far |
121 |
#endif
|
122 |
} FlashSV2Context; |
123 |
|
124 |
static av_cold void cleanup(FlashSV2Context * s) |
125 |
{ |
126 |
av_freep(&s->encbuffer); |
127 |
av_freep(&s->keybuffer); |
128 |
av_freep(&s->databuffer); |
129 |
av_freep(&s->current_frame); |
130 |
av_freep(&s->key_frame); |
131 |
|
132 |
av_freep(&s->frame_blocks); |
133 |
av_freep(&s->key_blocks); |
134 |
} |
135 |
|
136 |
static void init_blocks(FlashSV2Context * s, Block * blocks, |
137 |
uint8_t * encbuf, uint8_t * databuf) |
138 |
{ |
139 |
int row, col;
|
140 |
Block *b; |
141 |
for (col = 0; col < s->cols; col++) { |
142 |
for (row = 0; row < s->rows; row++) { |
143 |
b = blocks + (col + row * s->cols); |
144 |
b->width = (col < s->cols - 1) ?
|
145 |
s->block_width : |
146 |
s->image_width - col * s->block_width; |
147 |
|
148 |
b->height = (row < s->rows - 1) ?
|
149 |
s->block_height : |
150 |
s->image_height - row * s->block_height; |
151 |
|
152 |
b->row = row; |
153 |
b->col = col; |
154 |
b->enc = encbuf; |
155 |
b->data = databuf; |
156 |
encbuf += b->width * b->height * 3;
|
157 |
databuf += !databuf ? 0 : b->width * b->height * 6; |
158 |
} |
159 |
} |
160 |
} |
161 |
|
162 |
static void reset_stats(FlashSV2Context * s) |
163 |
{ |
164 |
#ifndef FLASHSV2_DUMB
|
165 |
s->diff_blocks = 0.1; |
166 |
s->tot_blocks = 1;
|
167 |
s->diff_lines = 0.1; |
168 |
s->tot_lines = 1;
|
169 |
s->raw_size = s->comp_size = s->uncomp_size = 10;
|
170 |
#endif
|
171 |
} |
172 |
|
173 |
static av_cold int flashsv2_encode_init(AVCodecContext * avctx) |
174 |
{ |
175 |
FlashSV2Context *s = avctx->priv_data; |
176 |
|
177 |
s->avctx = avctx; |
178 |
|
179 |
s->comp = avctx->compression_level; |
180 |
if (s->comp == -1) |
181 |
s->comp = 9;
|
182 |
if (s->comp < 0 || s->comp > 9) { |
183 |
av_log(avctx, AV_LOG_ERROR, |
184 |
"Compression level should be 0-9, not %d\n", s->comp);
|
185 |
return -1; |
186 |
} |
187 |
|
188 |
|
189 |
if ((avctx->width > 4095) || (avctx->height > 4095)) { |
190 |
av_log(avctx, AV_LOG_ERROR, |
191 |
"Input dimensions too large, input must be max 4096x4096 !\n");
|
192 |
return -1; |
193 |
} |
194 |
|
195 |
if (av_image_check_size(avctx->width, avctx->height, 0, avctx) < 0) |
196 |
return -1; |
197 |
|
198 |
|
199 |
s->last_key_frame = 0;
|
200 |
|
201 |
s->image_width = avctx->width; |
202 |
s->image_height = avctx->height; |
203 |
|
204 |
s->block_width = (s->image_width / 12) & ~15; |
205 |
s->block_height = (s->image_height / 12) & ~15; |
206 |
|
207 |
s->rows = (s->image_height + s->block_height - 1) / s->block_height;
|
208 |
s->cols = (s->image_width + s->block_width - 1) / s->block_width;
|
209 |
|
210 |
s->frame_size = s->image_width * s->image_height * 3;
|
211 |
s->blocks_size = s->rows * s->cols * sizeof(Block);
|
212 |
|
213 |
s->encbuffer = av_mallocz(s->frame_size); |
214 |
s->keybuffer = av_mallocz(s->frame_size); |
215 |
s->databuffer = av_mallocz(s->frame_size * 6);
|
216 |
s->current_frame = av_mallocz(s->frame_size); |
217 |
s->key_frame = av_mallocz(s->frame_size); |
218 |
s->frame_blocks = av_mallocz(s->blocks_size); |
219 |
s->key_blocks = av_mallocz(s->blocks_size); |
220 |
|
221 |
init_blocks(s, s->frame_blocks, s->encbuffer, s->databuffer); |
222 |
init_blocks(s, s->key_blocks, s->keybuffer, 0);
|
223 |
reset_stats(s); |
224 |
#ifndef FLASHSV2_DUMB
|
225 |
s->total_bits = 1;
|
226 |
#endif
|
227 |
|
228 |
s->use_custom_palette = 0;
|
229 |
s->palette_type = -1; // so that the palette will be generated in reconfigure_at_keyframe |
230 |
|
231 |
if (!s->encbuffer || !s->keybuffer || !s->databuffer
|
232 |
|| !s->current_frame || !s->key_frame || !s->key_blocks |
233 |
|| !s->frame_blocks) { |
234 |
av_log(avctx, AV_LOG_ERROR, "Memory allocation failed.\n");
|
235 |
cleanup(s); |
236 |
return -1; |
237 |
} |
238 |
|
239 |
return 0; |
240 |
} |
241 |
|
242 |
static int new_key_frame(FlashSV2Context * s) |
243 |
{ |
244 |
int i;
|
245 |
FFSWAP(uint8_t * , s->keybuffer, s->encbuffer); |
246 |
memcpy(s->key_blocks, s->frame_blocks, s->blocks_size); |
247 |
memcpy(s->key_frame, s->current_frame, s->frame_size); |
248 |
|
249 |
for (i = 0; i < s->rows * s->cols; i++) { |
250 |
s->key_blocks[i].enc += (s->keybuffer - s->encbuffer); |
251 |
s->key_blocks[i].sl_begin = 0;
|
252 |
s->key_blocks[i].sl_end = 0;
|
253 |
s->key_blocks[i].data = 0;
|
254 |
} |
255 |
|
256 |
return 0; |
257 |
} |
258 |
|
259 |
static int write_palette(FlashSV2Context * s, uint8_t * buf, int buf_size) |
260 |
{ |
261 |
//this isn't implemented yet! Default palette only!
|
262 |
return -1; |
263 |
} |
264 |
|
265 |
static int write_header(FlashSV2Context * s, uint8_t * buf, int buf_size) |
266 |
{ |
267 |
PutBitContext pb; |
268 |
int buf_pos, len;
|
269 |
|
270 |
if (buf_size < 5) |
271 |
return -1; |
272 |
|
273 |
init_put_bits(&pb, buf, buf_size * 8);
|
274 |
|
275 |
put_bits(&pb, 4, (s->block_width >> 4) - 1); |
276 |
put_bits(&pb, 12, s->image_width);
|
277 |
put_bits(&pb, 4, (s->block_height >> 4) - 1); |
278 |
put_bits(&pb, 12, s->image_height);
|
279 |
|
280 |
flush_put_bits(&pb); |
281 |
buf_pos = 4;
|
282 |
|
283 |
buf[buf_pos++] = s->flags; |
284 |
|
285 |
if (s->flags & HAS_PALLET_INFO) {
|
286 |
len = write_palette(s, buf + buf_pos, buf_size - buf_pos); |
287 |
if (len < 0) |
288 |
return -1; |
289 |
buf_pos += len; |
290 |
} |
291 |
|
292 |
return buf_pos;
|
293 |
} |
294 |
|
295 |
static int write_block(Block * b, uint8_t * buf, int buf_size) |
296 |
{ |
297 |
int buf_pos = 0; |
298 |
unsigned block_size = b->data_size;
|
299 |
|
300 |
if (b->flags & HAS_DIFF_BLOCKS)
|
301 |
block_size += 2;
|
302 |
if (b->flags & ZLIB_PRIME_COMPRESS_CURRENT)
|
303 |
block_size += 2;
|
304 |
if (block_size > 0) |
305 |
block_size += 1;
|
306 |
if (buf_size < block_size + 2) |
307 |
return -1; |
308 |
|
309 |
buf[buf_pos++] = block_size >> 8;
|
310 |
buf[buf_pos++] = block_size; |
311 |
|
312 |
if (block_size == 0) |
313 |
return buf_pos;
|
314 |
|
315 |
buf[buf_pos++] = b->flags; |
316 |
|
317 |
if (b->flags & HAS_DIFF_BLOCKS) {
|
318 |
buf[buf_pos++] = (uint8_t) (b->start); |
319 |
buf[buf_pos++] = (uint8_t) (b->len); |
320 |
} |
321 |
|
322 |
if (b->flags & ZLIB_PRIME_COMPRESS_CURRENT) {
|
323 |
//This feature of the format is poorly understood, and as of now, unused.
|
324 |
buf[buf_pos++] = (uint8_t) (b->col); |
325 |
buf[buf_pos++] = (uint8_t) (b->row); |
326 |
} |
327 |
|
328 |
memcpy(buf + buf_pos, b->data, b->data_size); |
329 |
|
330 |
buf_pos += b->data_size; |
331 |
|
332 |
return buf_pos;
|
333 |
} |
334 |
|
335 |
static int encode_zlib(Block * b, uint8_t * buf, unsigned long *buf_size, int comp) |
336 |
{ |
337 |
int res = compress2(buf, buf_size, b->sl_begin, b->sl_end - b->sl_begin, comp);
|
338 |
return res == Z_OK ? 0 : -1; |
339 |
} |
340 |
|
341 |
static int encode_zlibprime(Block * b, Block * prime, uint8_t * buf, |
342 |
int *buf_size, int comp) |
343 |
{ |
344 |
z_stream s; |
345 |
int res;
|
346 |
s.zalloc = NULL;
|
347 |
s.zfree = NULL;
|
348 |
s.opaque = NULL;
|
349 |
res = deflateInit(&s, comp); |
350 |
if (res < 0) |
351 |
return -1; |
352 |
|
353 |
s.next_in = prime->enc; |
354 |
s.avail_in = prime->enc_size; |
355 |
while (s.avail_in > 0) { |
356 |
s.next_out = buf; |
357 |
s.avail_out = *buf_size; |
358 |
res = deflate(&s, Z_SYNC_FLUSH); |
359 |
if (res < 0) |
360 |
return -1; |
361 |
} |
362 |
|
363 |
s.next_in = b->sl_begin; |
364 |
s.avail_in = b->sl_end - b->sl_begin; |
365 |
s.next_out = buf; |
366 |
s.avail_out = *buf_size; |
367 |
res = deflate(&s, Z_FINISH); |
368 |
deflateEnd(&s); |
369 |
*buf_size -= s.avail_out; |
370 |
if (res != Z_STREAM_END)
|
371 |
return -1; |
372 |
return 0; |
373 |
} |
374 |
|
375 |
static int encode_bgr(Block * b, const uint8_t * src, int stride) |
376 |
{ |
377 |
int i;
|
378 |
uint8_t *ptr = b->enc; |
379 |
for (i = 0; i < b->start; i++) |
380 |
memcpy(ptr + i * b->width * 3, src + i * stride, b->width * 3); |
381 |
b->sl_begin = ptr + i * b->width * 3;
|
382 |
for (; i < b->start + b->len; i++)
|
383 |
memcpy(ptr + i * b->width * 3, src + i * stride, b->width * 3); |
384 |
b->sl_end = ptr + i * b->width * 3;
|
385 |
for (; i < b->height; i++)
|
386 |
memcpy(ptr + i * b->width * 3, src + i * stride, b->width * 3); |
387 |
b->enc_size = ptr + i * b->width * 3 - b->enc;
|
388 |
return b->enc_size;
|
389 |
} |
390 |
|
391 |
static inline unsigned pixel_color15(const uint8_t * src) |
392 |
{ |
393 |
return (src[0] >> 3) | ((src[1] & 0xf8) << 2) | ((src[2] & 0xf8) << 7); |
394 |
} |
395 |
|
396 |
static inline unsigned int chroma_diff(unsigned int c1, unsigned int c2) |
397 |
{ |
398 |
unsigned int t1 = (c1 & 0x000000ff) + ((c1 & 0x0000ff00) >> 8) + ((c1 & 0x00ff0000) >> 16); |
399 |
unsigned int t2 = (c2 & 0x000000ff) + ((c2 & 0x0000ff00) >> 8) + ((c2 & 0x00ff0000) >> 16); |
400 |
|
401 |
return abs(t1 - t2) + abs((c1 & 0x000000ff) - (c2 & 0x000000ff)) + |
402 |
abs(((c1 & 0x0000ff00) >> 8) - ((c2 & 0x0000ff00) >> 8)) + |
403 |
abs(((c1 & 0x00ff0000) >> 16) - ((c2 & 0x00ff0000) >> 16)); |
404 |
} |
405 |
|
406 |
static inline int pixel_color7_fast(Palette * palette, unsigned c15) |
407 |
{ |
408 |
return palette->index[c15];
|
409 |
} |
410 |
|
411 |
static int pixel_color7_slow(Palette * palette, unsigned color) |
412 |
{ |
413 |
int i, min = 0x7fffffff; |
414 |
int minc = -1; |
415 |
for (i = 0; i < 128; i++) { |
416 |
int c1 = palette->colors[i];
|
417 |
int diff = chroma_diff(c1, color);
|
418 |
if (diff < min) {
|
419 |
min = diff; |
420 |
minc = i; |
421 |
} |
422 |
} |
423 |
return minc;
|
424 |
} |
425 |
|
426 |
static inline unsigned pixel_bgr(const uint8_t * src) |
427 |
{ |
428 |
return (src[0]) | (src[1] << 8) | (src[2] << 16); |
429 |
} |
430 |
|
431 |
static int write_pixel_15_7(Palette * palette, uint8_t * dest, const uint8_t * src, |
432 |
int dist)
|
433 |
{ |
434 |
unsigned c15 = pixel_color15(src);
|
435 |
unsigned color = pixel_bgr(src);
|
436 |
int d15 = chroma_diff(color, color & 0x00f8f8f8); |
437 |
int c7 = pixel_color7_fast(palette, c15);
|
438 |
int d7 = chroma_diff(color, palette->colors[c7]);
|
439 |
if (dist + d15 >= d7) {
|
440 |
dest[0] = (uint8_t) c7;
|
441 |
return 1; |
442 |
} else {
|
443 |
dest[0] = 0x80 | (uint8_t) (c15 >> 8); |
444 |
dest[1] = c15 & 0xff; |
445 |
return 2; |
446 |
} |
447 |
} |
448 |
|
449 |
static int update_palette_index(Palette * palette) |
450 |
{ |
451 |
int r, g, b;
|
452 |
unsigned int bgr, c15, index; |
453 |
for (r = 4; r < 256; r += 8) { |
454 |
for (g = 4; g < 256; g += 8) { |
455 |
for (b = 4; b < 256; b += 8) { |
456 |
bgr = b | (g << 8) | (r << 16); |
457 |
c15 = (b >> 3) | ((g & 0xf8) << 2) | ((r & 0xf8) << 7); |
458 |
index = pixel_color7_slow(palette, bgr); |
459 |
|
460 |
palette->index[c15] = index; |
461 |
} |
462 |
} |
463 |
} |
464 |
return 0; |
465 |
} |
466 |
|
467 |
static const unsigned int default_screen_video_v2_palette[128] = { |
468 |
0x00000000, 0x00333333, 0x00666666, 0x00999999, 0x00CCCCCC, 0x00FFFFFF, |
469 |
0x00330000, 0x00660000, 0x00990000, 0x00CC0000, 0x00FF0000, 0x00003300, |
470 |
0x00006600, 0x00009900, 0x0000CC00, 0x0000FF00, 0x00000033, 0x00000066, |
471 |
0x00000099, 0x000000CC, 0x000000FF, 0x00333300, 0x00666600, 0x00999900, |
472 |
0x00CCCC00, 0x00FFFF00, 0x00003333, 0x00006666, 0x00009999, 0x0000CCCC, |
473 |
0x0000FFFF, 0x00330033, 0x00660066, 0x00990099, 0x00CC00CC, 0x00FF00FF, |
474 |
0x00FFFF33, 0x00FFFF66, 0x00FFFF99, 0x00FFFFCC, 0x00FF33FF, 0x00FF66FF, |
475 |
0x00FF99FF, 0x00FFCCFF, 0x0033FFFF, 0x0066FFFF, 0x0099FFFF, 0x00CCFFFF, |
476 |
0x00CCCC33, 0x00CCCC66, 0x00CCCC99, 0x00CCCCFF, 0x00CC33CC, 0x00CC66CC, |
477 |
0x00CC99CC, 0x00CCFFCC, 0x0033CCCC, 0x0066CCCC, 0x0099CCCC, 0x00FFCCCC, |
478 |
0x00999933, 0x00999966, 0x009999CC, 0x009999FF, 0x00993399, 0x00996699, |
479 |
0x0099CC99, 0x0099FF99, 0x00339999, 0x00669999, 0x00CC9999, 0x00FF9999, |
480 |
0x00666633, 0x00666699, 0x006666CC, 0x006666FF, 0x00663366, 0x00669966, |
481 |
0x0066CC66, 0x0066FF66, 0x00336666, 0x00996666, 0x00CC6666, 0x00FF6666, |
482 |
0x00333366, 0x00333399, 0x003333CC, 0x003333FF, 0x00336633, 0x00339933, |
483 |
0x0033CC33, 0x0033FF33, 0x00663333, 0x00993333, 0x00CC3333, 0x00FF3333, |
484 |
0x00003366, 0x00336600, 0x00660033, 0x00006633, 0x00330066, 0x00663300, |
485 |
0x00336699, 0x00669933, 0x00993366, 0x00339966, 0x00663399, 0x00996633, |
486 |
0x006699CC, 0x0099CC66, 0x00CC6699, 0x0066CC99, 0x009966CC, 0x00CC9966, |
487 |
0x0099CCFF, 0x00CCFF99, 0x00FF99CC, 0x0099FFCC, 0x00CC99FF, 0x00FFCC99, |
488 |
0x00111111, 0x00222222, 0x00444444, 0x00555555, 0x00AAAAAA, 0x00BBBBBB, |
489 |
0x00DDDDDD, 0x00EEEEEE |
490 |
}; |
491 |
|
492 |
static int generate_default_palette(Palette * palette) |
493 |
{ |
494 |
memcpy(palette->colors, default_screen_video_v2_palette, |
495 |
sizeof(default_screen_video_v2_palette));
|
496 |
|
497 |
return update_palette_index(palette);
|
498 |
} |
499 |
|
500 |
static int generate_optimum_palette(Palette * palette, const uint8_t * image, |
501 |
int width, int height, int stride) |
502 |
{ |
503 |
//this isn't implemented yet! Default palette only!
|
504 |
return -1; |
505 |
} |
506 |
|
507 |
static inline int encode_15_7_sl(Palette * palette, uint8_t * dest, |
508 |
const uint8_t * src, int width, int dist) |
509 |
{ |
510 |
int len = 0, x; |
511 |
for (x = 0; x < width; x++) { |
512 |
len += write_pixel_15_7(palette, dest + len, src + 3 * x, dist);
|
513 |
} |
514 |
return len;
|
515 |
} |
516 |
|
517 |
static int encode_15_7(Palette * palette, Block * b, const uint8_t * src, |
518 |
int stride, int dist) |
519 |
{ |
520 |
int i;
|
521 |
uint8_t *ptr = b->enc; |
522 |
for (i = 0; i < b->start; i++) |
523 |
ptr += encode_15_7_sl(palette, ptr, src + i * stride, b->width, dist); |
524 |
b->sl_begin = ptr; |
525 |
for (; i < b->start + b->len; i++)
|
526 |
ptr += encode_15_7_sl(palette, ptr, src + i * stride, b->width, dist); |
527 |
b->sl_end = ptr; |
528 |
for (; i < b->height; i++)
|
529 |
ptr += encode_15_7_sl(palette, ptr, src + i * stride, b->width, dist); |
530 |
b->enc_size = ptr - b->enc; |
531 |
return b->enc_size;
|
532 |
} |
533 |
|
534 |
static int encode_block(Palette * palette, Block * b, Block * prev, |
535 |
const uint8_t * src, int stride, int comp, int dist, |
536 |
int keyframe)
|
537 |
{ |
538 |
unsigned buf_size = b->width * b->height * 6; |
539 |
uint8_t buf[buf_size]; |
540 |
int res;
|
541 |
if (b->flags & COLORSPACE_15_7) {
|
542 |
encode_15_7(palette, b, src, stride, dist); |
543 |
} else {
|
544 |
encode_bgr(b, src, stride); |
545 |
} |
546 |
|
547 |
if (b->len > 0) { |
548 |
b->data_size = buf_size; |
549 |
res = encode_zlib(b, b->data, &b->data_size, comp); |
550 |
if (res)
|
551 |
return res;
|
552 |
|
553 |
if (!keyframe) {
|
554 |
res = encode_zlibprime(b, prev, buf, &buf_size, comp); |
555 |
if (res)
|
556 |
return res;
|
557 |
|
558 |
if (buf_size < b->data_size) {
|
559 |
b->data_size = buf_size; |
560 |
memcpy(b->data, buf, buf_size); |
561 |
b->flags |= ZLIB_PRIME_COMPRESS_PREVIOUS; |
562 |
} |
563 |
} |
564 |
} else {
|
565 |
b->data_size = 0;
|
566 |
} |
567 |
return 0; |
568 |
} |
569 |
|
570 |
static int compare_sl(FlashSV2Context * s, Block * b, const uint8_t * src, |
571 |
uint8_t * frame, uint8_t * key, int y, int keyframe) |
572 |
{ |
573 |
if (memcmp(src, frame, b->width * 3) != 0) { |
574 |
b->dirty = 1;
|
575 |
memcpy(frame, src, b->width * 3);
|
576 |
#ifndef FLASHSV2_DUMB
|
577 |
s->diff_lines++; |
578 |
#endif
|
579 |
} |
580 |
if (memcmp(src, key, b->width * 3) != 0) { |
581 |
if (b->len == 0) |
582 |
b->start = y; |
583 |
b->len = y + 1 - b->start;
|
584 |
} |
585 |
return 0; |
586 |
} |
587 |
|
588 |
static int mark_all_blocks(FlashSV2Context * s, const uint8_t * src, int stride, |
589 |
int keyframe)
|
590 |
{ |
591 |
int sl, rsl, col, pos, possl;
|
592 |
Block *b; |
593 |
for (sl = s->image_height - 1; sl >= 0; sl--) { |
594 |
for (col = 0; col < s->cols; col++) { |
595 |
rsl = s->image_height - sl - 1;
|
596 |
b = s->frame_blocks + col + rsl / s->block_height * s->cols; |
597 |
possl = stride * sl + col * s->block_width * 3;
|
598 |
pos = s->image_width * rsl * 3 + col * s->block_width * 3; |
599 |
compare_sl(s, b, src + possl, s->current_frame + pos, |
600 |
s->key_frame + pos, rsl % s->block_height, keyframe); |
601 |
} |
602 |
} |
603 |
#ifndef FLASHSV2_DUMB
|
604 |
s->tot_lines += s->image_height * s->cols; |
605 |
#endif
|
606 |
return 0; |
607 |
} |
608 |
|
609 |
static int encode_all_blocks(FlashSV2Context * s, int keyframe) |
610 |
{ |
611 |
int row, col, res;
|
612 |
uint8_t *data; |
613 |
Block *b, *prev; |
614 |
for (row = 0; row < s->rows; row++) { |
615 |
for (col = 0; col < s->cols; col++) { |
616 |
b = s->frame_blocks + (row * s->cols + col); |
617 |
prev = s->key_blocks + (row * s->cols + col); |
618 |
if (keyframe) {
|
619 |
b->start = 0;
|
620 |
b->len = b->height; |
621 |
b->flags = s->use15_7 ? COLORSPACE_15_7 : 0;
|
622 |
} else if (!b->dirty) { |
623 |
b->start = 0;
|
624 |
b->len = 0;
|
625 |
b->data_size = 0;
|
626 |
b->flags = s->use15_7 ? COLORSPACE_15_7 : 0;
|
627 |
continue;
|
628 |
} else {
|
629 |
b->flags = s->use15_7 ? COLORSPACE_15_7 | HAS_DIFF_BLOCKS : HAS_DIFF_BLOCKS; |
630 |
} |
631 |
data = s->current_frame + s->image_width * 3 * s->block_height * row + s->block_width * col * 3; |
632 |
res = encode_block(&s->palette, b, prev, data, s->image_width * 3, s->comp, s->dist, keyframe);
|
633 |
#ifndef FLASHSV2_DUMB
|
634 |
if (b->dirty)
|
635 |
s->diff_blocks++; |
636 |
s->comp_size += b->data_size; |
637 |
s->uncomp_size += b->enc_size; |
638 |
#endif
|
639 |
if (res)
|
640 |
return res;
|
641 |
} |
642 |
} |
643 |
#ifndef FLASHSV2_DUMB
|
644 |
s->raw_size += s->image_width * s->image_height * 3;
|
645 |
s->tot_blocks += s->rows * s->cols; |
646 |
#endif
|
647 |
return 0; |
648 |
} |
649 |
|
650 |
static int write_all_blocks(FlashSV2Context * s, uint8_t * buf, |
651 |
int buf_size)
|
652 |
{ |
653 |
int row, col, buf_pos = 0, len; |
654 |
Block *b; |
655 |
for (row = 0; row < s->rows; row++) { |
656 |
for (col = 0; col < s->cols; col++) { |
657 |
b = s->frame_blocks + row * s->cols + col; |
658 |
len = write_block(b, buf + buf_pos, buf_size - buf_pos); |
659 |
b->start = b->len = b->dirty = 0;
|
660 |
if (len < 0) |
661 |
return len;
|
662 |
buf_pos += len; |
663 |
} |
664 |
} |
665 |
return buf_pos;
|
666 |
} |
667 |
|
668 |
static int write_bitstream(FlashSV2Context * s, const uint8_t * src, int stride, |
669 |
uint8_t * buf, int buf_size, int keyframe) |
670 |
{ |
671 |
int buf_pos, res;
|
672 |
|
673 |
res = mark_all_blocks(s, src, stride, keyframe); |
674 |
if (res)
|
675 |
return res;
|
676 |
res = encode_all_blocks(s, keyframe); |
677 |
if (res)
|
678 |
return res;
|
679 |
|
680 |
res = write_header(s, buf, buf_size); |
681 |
if (res < 0) { |
682 |
return res;
|
683 |
} else {
|
684 |
buf_pos = res; |
685 |
} |
686 |
res = write_all_blocks(s, buf + buf_pos, buf_size - buf_pos); |
687 |
if (res < 0) |
688 |
return res;
|
689 |
buf_pos += res; |
690 |
#ifndef FLASHSV2_DUMB
|
691 |
s->total_bits += ((double) buf_pos) * 8.0; |
692 |
#endif
|
693 |
|
694 |
return buf_pos;
|
695 |
} |
696 |
|
697 |
static void recommend_keyframe(FlashSV2Context * s, int *keyframe) |
698 |
{ |
699 |
#ifndef FLASHSV2_DUMB
|
700 |
double block_ratio, line_ratio, enc_ratio, comp_ratio, data_ratio;
|
701 |
if (s->avctx->gop_size > 0) { |
702 |
block_ratio = s->diff_blocks / s->tot_blocks; |
703 |
line_ratio = s->diff_lines / s->tot_lines; |
704 |
enc_ratio = s->uncomp_size / s->raw_size; |
705 |
comp_ratio = s->comp_size / s->uncomp_size; |
706 |
data_ratio = s->comp_size / s->raw_size; |
707 |
|
708 |
if ((block_ratio >= 0.5 && line_ratio / block_ratio <= 0.5) || line_ratio >= 0.95) { |
709 |
*keyframe = 1;
|
710 |
return;
|
711 |
} |
712 |
} |
713 |
#else
|
714 |
return;
|
715 |
#endif
|
716 |
} |
717 |
|
718 |
static const double block_size_fraction = 1.0 / 300; |
719 |
static int optimum_block_width(FlashSV2Context * s) |
720 |
{ |
721 |
#ifndef FLASHSV2_DUMB
|
722 |
double save = (1-pow(s->diff_lines/s->diff_blocks/s->block_height, 0.5)) * s->comp_size/s->tot_blocks; |
723 |
double width = block_size_fraction * sqrt(0.5 * save * s->rows * s->cols) * s->image_width; |
724 |
int pwidth = ((int) width); |
725 |
return FFCLIP(pwidth & ~15, 256, 16); |
726 |
#else
|
727 |
return 64; |
728 |
#endif
|
729 |
} |
730 |
|
731 |
static int optimum_block_height(FlashSV2Context * s) |
732 |
{ |
733 |
#ifndef FLASHSV2_DUMB
|
734 |
double save = (1-pow(s->diff_lines/s->diff_blocks/s->block_height, 0.5)) * s->comp_size/s->tot_blocks; |
735 |
double height = block_size_fraction * sqrt(0.5 * save * s->rows * s->cols) * s->image_height; |
736 |
int pheight = ((int) height); |
737 |
return FFCLIP(pheight & ~15, 256, 16); |
738 |
#else
|
739 |
return 64; |
740 |
#endif
|
741 |
} |
742 |
|
743 |
static const double use15_7_threshold = 8192; |
744 |
|
745 |
static int optimum_use15_7(FlashSV2Context * s) |
746 |
{ |
747 |
#ifndef FLASHSV2_DUMB
|
748 |
double ideal = ((double)(s->avctx->bit_rate * s->avctx->time_base.den * s->avctx->ticks_per_frame)) / |
749 |
((double) s->avctx->time_base.num) * s->avctx->frame_number;
|
750 |
if (ideal + use15_7_threshold < s->total_bits) {
|
751 |
return 1; |
752 |
} else {
|
753 |
return 0; |
754 |
} |
755 |
#else
|
756 |
return s->avctx->global_quality == 0; |
757 |
#endif
|
758 |
} |
759 |
|
760 |
static const double color15_7_factor = 100; |
761 |
|
762 |
static int optimum_dist(FlashSV2Context * s) |
763 |
{ |
764 |
#ifndef FLASHSV2_DUMB
|
765 |
double ideal =
|
766 |
s->avctx->bit_rate * s->avctx->time_base.den * |
767 |
s->avctx->ticks_per_frame; |
768 |
int dist = pow((s->total_bits / ideal) * color15_7_factor, 3); |
769 |
av_log(s->avctx, AV_LOG_DEBUG, "dist: %d\n", dist);
|
770 |
return dist;
|
771 |
#else
|
772 |
return 15; |
773 |
#endif
|
774 |
} |
775 |
|
776 |
|
777 |
static int reconfigure_at_keyframe(FlashSV2Context * s, const uint8_t * image, |
778 |
int stride)
|
779 |
{ |
780 |
int update_palette = 0; |
781 |
int res;
|
782 |
s->block_width = optimum_block_width(s); |
783 |
s->block_height = optimum_block_height(s); |
784 |
|
785 |
s->rows = (s->image_height + s->block_height - 1) / s->block_height;
|
786 |
s->cols = (s->image_width + s->block_width - 1) / s->block_width;
|
787 |
|
788 |
if (s->rows * s->cols != s->blocks_size / sizeof(Block)) { |
789 |
if (s->rows * s->cols > s->blocks_size / sizeof(Block)) { |
790 |
s->frame_blocks = av_realloc(s->frame_blocks, s->rows * s->cols * sizeof(Block));
|
791 |
s->key_blocks = av_realloc(s->key_blocks, s->cols * s->rows * sizeof(Block));
|
792 |
if (!s->frame_blocks || !s->key_blocks) {
|
793 |
av_log(s->avctx, AV_LOG_ERROR, "Memory allocation failed.\n");
|
794 |
return -1; |
795 |
} |
796 |
s->blocks_size = s->rows * s->cols * sizeof(Block);
|
797 |
} |
798 |
init_blocks(s, s->frame_blocks, s->encbuffer, s->databuffer); |
799 |
init_blocks(s, s->key_blocks, s->keybuffer, 0);
|
800 |
|
801 |
} |
802 |
|
803 |
s->use15_7 = optimum_use15_7(s); |
804 |
if (s->use15_7) {
|
805 |
if ((s->use_custom_palette && s->palette_type != 1) || update_palette) { |
806 |
res = generate_optimum_palette(&s->palette, image, s->image_width, s->image_height, stride); |
807 |
if (res)
|
808 |
return res;
|
809 |
s->palette_type = 1;
|
810 |
av_log(s->avctx, AV_LOG_DEBUG, "Generated optimum palette\n");
|
811 |
} else if (!s->use_custom_palette && s->palette_type != 0) { |
812 |
res = generate_default_palette(&s->palette); |
813 |
if (res)
|
814 |
return res;
|
815 |
s->palette_type = 0;
|
816 |
av_log(s->avctx, AV_LOG_DEBUG, "Generated default palette\n");
|
817 |
} |
818 |
} |
819 |
|
820 |
|
821 |
reset_stats(s); |
822 |
|
823 |
return 0; |
824 |
} |
825 |
|
826 |
static int flashsv2_encode_frame(AVCodecContext * avctx, uint8_t * buf, |
827 |
int buf_size, void *data) |
828 |
{ |
829 |
FlashSV2Context *const s = avctx->priv_data;
|
830 |
AVFrame *pict = data; |
831 |
AVFrame *const p = &s->frame;
|
832 |
int res;
|
833 |
int keyframe = 0; |
834 |
|
835 |
*p = *pict; |
836 |
|
837 |
/* First frame needs to be a keyframe */
|
838 |
if (avctx->frame_number == 0) |
839 |
keyframe = 1;
|
840 |
|
841 |
/* Check the placement of keyframes */
|
842 |
if (avctx->gop_size > 0) { |
843 |
if (avctx->frame_number >= s->last_key_frame + avctx->gop_size)
|
844 |
keyframe = 1;
|
845 |
} |
846 |
|
847 |
if (buf_size < s->frame_size) {
|
848 |
//Conservative upper bound check for compressed data
|
849 |
av_log(avctx, AV_LOG_ERROR, "buf_size %d < %d\n", buf_size, s->frame_size);
|
850 |
return -1; |
851 |
} |
852 |
|
853 |
if (!keyframe
|
854 |
&& avctx->frame_number > s->last_key_frame + avctx->keyint_min) { |
855 |
recommend_keyframe(s, &keyframe); |
856 |
if (keyframe)
|
857 |
av_log(avctx, AV_LOG_DEBUG, "Recommending key frame at frame %d\n", avctx->frame_number);
|
858 |
} |
859 |
|
860 |
if (keyframe) {
|
861 |
res = reconfigure_at_keyframe(s, p->data[0], p->linesize[0]); |
862 |
if (res)
|
863 |
return res;
|
864 |
} |
865 |
|
866 |
if (s->use15_7)
|
867 |
s->dist = optimum_dist(s); |
868 |
|
869 |
res = write_bitstream(s, p->data[0], p->linesize[0], buf, buf_size, keyframe); |
870 |
|
871 |
if (keyframe) {
|
872 |
new_key_frame(s); |
873 |
p->pict_type = FF_I_TYPE; |
874 |
p->key_frame = 1;
|
875 |
s->last_key_frame = avctx->frame_number; |
876 |
av_log(avctx, AV_LOG_DEBUG, "Inserting key frame at frame %d\n", avctx->frame_number);
|
877 |
} else {
|
878 |
p->pict_type = FF_P_TYPE; |
879 |
p->key_frame = 0;
|
880 |
} |
881 |
|
882 |
avctx->coded_frame = p; |
883 |
|
884 |
return res;
|
885 |
} |
886 |
|
887 |
static av_cold int flashsv2_encode_end(AVCodecContext * avctx) |
888 |
{ |
889 |
FlashSV2Context *s = avctx->priv_data; |
890 |
|
891 |
cleanup(s); |
892 |
|
893 |
return 0; |
894 |
} |
895 |
|
896 |
AVCodec ff_flashsv2_encoder = { |
897 |
"flashsv2",
|
898 |
AVMEDIA_TYPE_VIDEO, |
899 |
CODEC_ID_FLASHSV2, |
900 |
sizeof(FlashSV2Context),
|
901 |
flashsv2_encode_init, |
902 |
flashsv2_encode_frame, |
903 |
flashsv2_encode_end, |
904 |
.pix_fmts = (enum PixelFormat[]) {PIX_FMT_BGR24, PIX_FMT_NONE},
|
905 |
.long_name = NULL_IF_CONFIG_SMALL("Flash Screen Video Version 2"),
|
906 |
}; |