1 /* swfvideo.c
2    Routines for handling h.263 video tags
3 
4    Part of the swftools package.
5 
6    Copyright (c) 2003 Matthias Kramm <kramm@quiss.org>
7 
8    This program is free software; you can redistribute it and/or modify
9    it under the terms of the GNU General Public License as published by
10    the Free Software Foundation; either version 2 of the License, or
11    (at your option) any later version.
12 
13    This program is distributed in the hope that it will be useful,
14    but WITHOUT ANY WARRANTY; without even the implied warranty of
15    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
16    GNU General Public License for more details.
17 
18    You should have received a copy of the GNU General Public License
19    along with this program; if not, write to the Free Software
20    Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA */
21 
22 #include <stdlib.h>
23 #include <stdio.h>
24 #include <assert.h>
25 #include <math.h>
26 #include "../rfxswf.h"
27 #include "h263tables.h"
28 #include "dct.h"
29 
30 /* TODO:
31    - use prepare* / write* in encode_IFrame_block
32    - check whether mvd steps of 2 lead to (much) smaller results
33 */
34 
35 #ifdef MAIN
36 U16 totalframes = 0;
37 #endif
swf_SetVideoStreamDefine(TAG * tag,VIDEOSTREAM * stream,U16 frames,U16 width,U16 height)38 void swf_SetVideoStreamDefine(TAG*tag, VIDEOSTREAM*stream, U16 frames, U16 width, U16 height)
39 {
40 #ifdef MAIN
41     totalframes = frames;
42 #endif
43     memset(stream, 0, sizeof(VIDEOSTREAM));
44     stream->olinex = width;
45     stream->owidth = width;
46     stream->oheight = height;
47     width+=15;width&=~15;
48     height+=15;height&=~15;
49     stream->linex = width;
50     stream->width = width;
51     stream->height = height;
52     stream->bbx = width/16;
53     stream->bby = height/16;
54     stream->current = (YUV*)rfx_calloc(width*height*sizeof(YUV));
55     stream->oldpic = (YUV*)rfx_calloc(width*height*sizeof(YUV));
56     stream->mvdx = (int*)rfx_alloc(stream->bbx*stream->bby*sizeof(int));
57     stream->mvdy = (int*)rfx_alloc(stream->bbx*stream->bby*sizeof(int));
58     stream->do_motion = 0;
59 
60     assert((stream->width&15) == 0);
61     assert((stream->height&15) == 0);
62     assert((stream->bbx*16) == stream->width);
63     assert((stream->bby*16) == stream->height);
64 
65     swf_SetU16(tag, frames);
66     swf_SetU16(tag, width);
67     swf_SetU16(tag, height);
68     //swf_SetU8(tag, 1); /* smoothing on */
69     swf_SetU8(tag, 0); /* smoothing off */
70     swf_SetU8(tag, 2); /* codec = h.263 sorenson spark */
71 
72 }
swf_VideoStreamClear(VIDEOSTREAM * stream)73 void swf_VideoStreamClear(VIDEOSTREAM*stream)
74 {
75     rfx_free(stream->oldpic);stream->oldpic = 0;
76     rfx_free(stream->current);stream->current = 0;
77     rfx_free(stream->mvdx);stream->mvdx=0;
78     rfx_free(stream->mvdy);stream->mvdy=0;
79 }
80 
81 typedef struct _block_t
82 {
83     int y1[64];
84     int y2[64];
85     int y3[64];
86     int y4[64];
87     int u[64];
88     int v[64];
89 } block_t;
90 
truncate256(int a)91 static inline int truncate256(int a)
92 {
93     if(a>255) return 255;
94     if(a<0) return 0;
95     return a;
96 }
97 
getregion(block_t * bb,YUV * pic,int posx,int posy,int linex)98 static void getregion(block_t* bb, YUV*pic, int posx, int posy, int linex)
99 {
100     YUV*p1;
101     YUV*p2;
102     int i=0;
103     int x,y;
104     posx*=16;
105     posy*=16;
106     p1 = &pic[posy*linex+posx];
107     p2 = p1;
108     for(y=0;y<8;y++) {
109 	for(x=0;x<8;x++) {
110 	    bb->u[i] = (p2[x*2].u + p2[x*2+1].u + p2[linex+x*2].u + p2[linex+x*2+1].u)/4;
111 	    bb->v[i] = (p2[x*2].v + p2[x*2+1].v + p2[linex+x*2].v + p2[linex+x*2+1].v)/4;
112 	    bb->y1[i] = p1[x].y;
113 	    bb->y2[i] = p1[x+8].y;
114 	    bb->y3[i] = p1[linex*8+x].y;
115 	    bb->y4[i] = p1[linex*8+x+8].y;
116 	    i++;
117 	}
118 	p1+=linex;
119 	p2+=linex*2;
120     }
121 }
122 
123 /* This function is pretty complex. Let's hope it works correctly */
getmvdregion(block_t * bb,YUV * pic,int posx,int posy,int mvdx,int mvdy,int linex)124 static void getmvdregion(block_t* bb, YUV*pic, int posx, int posy, int mvdx, int mvdy, int linex)
125 {
126     YUV*p1;
127     YUV*p2;
128     int yy=0,uv=0;
129     int x,y;
130     int yhp = 0, uvhp=0;
131     int uvposx, uvposy;
132     posx = posx*16 + ((mvdx&~1)/2); //works also for negative mvdx (unlike mvdx/2)
133     posy = posy*16 + ((mvdy&~1)/2);
134     p1 = &pic[posy*linex+posx];
135     p2 = &pic[(posy&~1)*linex+(posx&~1)];
136     uvhp = ((mvdx&1)|((mvdx>>1)&1))|((mvdy&2)|((mvdy&1)<<1));
137     yhp = ((mvdy&1)<<1|(mvdx&1));
138 
139     /* y */
140     if(yhp==0 || yhp==2) {
141 	for(y=0;y<8;y++) {
142 	    for(x=0;x<8;x++) {
143 		bb->y1[yy] = p1[x].y;
144 		bb->y2[yy] = p1[x+8].y;
145 		bb->y3[yy] = p1[linex*8+x].y;
146 		bb->y4[yy] = p1[linex*8+x+8].y;
147 		yy++;
148 	    }
149 	    p1+=linex;
150 
151 	    if(yhp==2) {
152 		yy-=8;
153 		for(x=0;x<8;x++) {
154 		    bb->y1[yy] += p1[x].y; bb->y1[yy] /= 2;
155 		    bb->y2[yy] += p1[x+8].y; bb->y2[yy] /= 2;
156 		    bb->y3[yy] += p1[linex*8+x].y; bb->y3[yy] /= 2;
157 		    bb->y4[yy] += p1[linex*8+x+8].y; bb->y4[yy] /= 2;
158 		    yy++;
159 		}
160 	    }
161 	}
162     } else if(yhp==1 || yhp==3) {
163 	for(y=0;y<8;y++) {
164 	    for(x=0;x<8;x++) {
165 		bb->y1[yy] = (p1[x].y + p1[x+1].y);
166 		bb->y2[yy] = (p1[x+8].y + p1[x+8+1].y);
167 		bb->y3[yy] = (p1[linex*8+x].y + p1[linex*8+x+1].y);
168 		bb->y4[yy] = (p1[linex*8+x+8].y + p1[linex*8+x+8+1].y);
169 		yy++;
170 	    }
171 	    yy-=8;
172 	    p1+=linex;
173 	    if(yhp==3) {
174 		for(x=0;x<8;x++) {
175 		    bb->y1[yy] += (p1[x].y + p1[x+1].y); bb->y1[yy]/=4;
176 		    bb->y2[yy] += (p1[x+8].y + p1[x+8+1].y); bb->y2[yy]/=4;
177 		    bb->y3[yy] += (p1[linex*8+x].y + p1[linex*8+x+1].y); bb->y3[yy]/=4;
178 		    bb->y4[yy] += (p1[linex*8+x+8].y + p1[linex*8+x+8+1].y); bb->y4[yy]/=4;
179 		    yy++;
180 		}
181 	    } else {
182 		for(x=0;x<8;x++) {
183 		    bb->y1[yy]/=2; bb->y2[yy]/=2; bb->y3[yy]/=2; bb->y4[yy]/=2;
184 		    yy++;
185 		}
186 	    }
187 	}
188     }
189 
190     /* u,v */
191     if(uvhp==0 || uvhp==2) {
192 	for(y=0;y<8;y++) {
193 	    for(x=0;x<8;x++) {
194 		bb->u[uv] = (p2[x*2].u + p2[x*2+1].u + p2[linex+x*2].u + p2[linex+x*2+1].u)/4;
195 		bb->v[uv] = (p2[x*2].v + p2[x*2+1].v + p2[linex+x*2].v + p2[linex+x*2+1].v)/4;
196 		uv++;
197 	    }
198 	    p2+=linex*2;
199 	    if(uvhp==2) {
200 		uv-=8;
201 		for(x=0;x<8;x++) {
202 		    bb->u[uv] += (p2[x*2].u + p2[x*2+1].u + p2[linex+x*2].u + p2[linex+x*2+1].u)/4;
203 		    bb->v[uv] += (p2[x*2].v + p2[x*2+1].v + p2[linex+x*2].v + p2[linex+x*2+1].v)/4;
204 		    bb->u[uv] /= 2;
205 		    bb->v[uv] /= 2;
206 		    uv++;
207 		}
208 	    }
209 	}
210     } else /* uvhp==1 || uvhp==3 */ {
211 	for(y=0;y<8;y++) {
212 	    for(x=0;x<8;x++) {
213 		bb->u[uv] = ((p2[x*2].u + p2[x*2+1].u + p2[linex+x*2].u + p2[linex+x*2+1].u)/4+
214 		             (p2[x*2+2].u + p2[x*2+1+2].u + p2[linex+x*2+2].u + p2[linex+x*2+1+2].u)/4);
215 		bb->v[uv] = ((p2[x*2].v + p2[x*2+1].v + p2[linex+x*2].v + p2[linex+x*2+1].v)/4+
216 			     (p2[x*2+2].v + p2[x*2+1+2].v + p2[linex+x*2+2].v + p2[linex+x*2+1+2].v)/4);
217 		uv++;
218 	    }
219 	    uv-=8;
220 	    p2+=linex*2;
221 	    if(uvhp==3) {
222 		for(x=0;x<8;x++) {
223 		    bb->u[uv] += ((p2[x*2].u + p2[x*2+1].u + p2[linex+x*2].u + p2[linex+x*2+1].u)/4+
224 			 	  (p2[x*2+2].u + p2[x*2+1+2].u + p2[linex+x*2+2].u + p2[linex+x*2+1+2].u)/4);
225 		    bb->v[uv] += ((p2[x*2].v + p2[x*2+1].v + p2[linex+x*2].v + p2[linex+x*2+1].v)/4+
226 				  (p2[x*2+2].v + p2[x*2+1+2].v + p2[linex+x*2+2].v + p2[linex+x*2+1+2].v)/4);
227 		    bb->u[uv] /= 4;
228 		    bb->v[uv] /= 4;
229 		    uv++;
230 		}
231 	    } else {
232 		for(x=0;x<8;x++) {
233 		    bb->u[uv] /= 2;
234 		    bb->v[uv] /= 2;
235 		    uv++;
236 		}
237 	    }
238 	}
239     }
240 }
241 
rgb2yuv(YUV * dest,RGBA * src,int dlinex,int slinex,int width,int height)242 static void rgb2yuv(YUV*dest, RGBA*src, int dlinex, int slinex, int width, int height)
243 {
244     int x,y;
245     for(y=0;y<height;y++) {
246 	for(x=0;x<width;x++) {
247 	    int r,g,b;
248 	    r = src[y*slinex+x].r;
249 	    g = src[y*slinex+x].g;
250 	    b = src[y*slinex+x].b;
251 	    /*dest[y*dlinex+x].y = (r*0.299 + g*0.587 + b*0.114);
252 	    dest[y*dlinex+x].u = (r*-0.169 + g*-0.332 + b*0.500 + 128.0);
253 	    dest[y*dlinex+x].v = (r*0.500 + g*-0.419 + b*-0.0813 + 128.0);*/
254 
255 	    //dest[y*dlinex+x].y = 128;//(r*((int)( 0.299*256)) + g*((int)( 0.587*256)) + b*((int)( 0.114 *256)))>>8;
256 
257 	    dest[y*dlinex+x].y = (r*((int)( 0.299*256)) + g*((int)( 0.587*256)) + b*((int)( 0.114 *256)))>>8;
258 	    dest[y*dlinex+x].u = (r*((int)(-0.169*256)) + g*((int)(-0.332*256)) + b*((int)( 0.500 *256))+ 128*256)>>8;
259 	    dest[y*dlinex+x].v = (r*((int)( 0.500*256)) + g*((int)(-0.419*256)) + b*((int)(-0.0813*256))+ 128*256)>>8;
260 	}
261     }
262 }
263 
copyregion(VIDEOSTREAM * s,YUV * dest,YUV * src,int bx,int by)264 static void copyregion(VIDEOSTREAM*s, YUV*dest, YUV*src, int bx, int by)
265 {
266     YUV*p1 = &dest[by*s->linex*16+bx*16];
267     YUV*p2 = &src[by*s->linex*16+bx*16];
268     int y;
269     for(y=0;y<16;y++) {
270 	memcpy(p1, p2, 16*sizeof(YUV));
271 	p1+=s->linex;p2+=s->linex;
272     }
273 }
274 
yuv2rgb(RGBA * dest,YUV * src,int linex,int width,int height)275 static void yuv2rgb(RGBA*dest, YUV*src, int linex, int width, int height)
276 {
277     int x,y;
278     for(y=0;y<height;y++) {
279 	for(x=0;x<width;x++) {
280 	    int u,v,yy;
281 	    u = src[y*linex+x].u;
282 	    v = src[y*linex+x].v;
283 	    yy = src[y*linex+x].y;
284 	    dest[y*linex+x].r = truncate256(yy + ((360*(v-128))>>8));
285 	    dest[y*linex+x].g = truncate256(yy - ((88*(u-128)+183*(v-128))>>8));
286 	    dest[y*linex+x].b = truncate256(yy + ((455 * (u-128))>>8));
287 	}
288     }
289 }
copy_block_pic(VIDEOSTREAM * s,YUV * dest,block_t * b,int bx,int by)290 static void copy_block_pic(VIDEOSTREAM*s, YUV*dest, block_t*b, int bx, int by)
291 {
292     YUV*p1 = &dest[(by*16)*s->linex+bx*16];
293     YUV*p2 = &dest[(by*16+8)*s->linex+bx*16];
294     int x,y;
295     for(y=0;y<8;y++) {
296 	for(x=0;x<8;x++) {
297 	    int u,v,yy;
298 	    p1[x+0].u = b->u[(y/2)*8+(x/2)];
299 	    p1[x+0].v = b->v[(y/2)*8+(x/2)];
300 	    p1[x+0].y = b->y1[y*8+x];
301 	    p1[x+8].u = b->u[(y/2)*8+(x/2)+4];
302 	    p1[x+8].v = b->v[(y/2)*8+(x/2)+4];
303 	    p1[x+8].y = b->y2[y*8+x];
304 	    p2[x+0].u = b->u[(y/2+4)*8+(x/2)];
305 	    p2[x+0].v = b->v[(y/2+4)*8+(x/2)];
306 	    p2[x+0].y = b->y3[y*8+x];
307 	    p2[x+8].u = b->u[(y/2+4)*8+(x/2)+4];
308 	    p2[x+8].v = b->v[(y/2+4)*8+(x/2)+4];
309 	    p2[x+8].y = b->y4[y*8+x];
310 	}
311 	p1+=s->linex;
312 	p2+=s->linex;
313     }
314 }
315 
compare_pic_pic(VIDEOSTREAM * s,YUV * pp1,YUV * pp2,int bx,int by)316 static int compare_pic_pic(VIDEOSTREAM*s, YUV*pp1, YUV*pp2, int bx, int by)
317 {
318     int linex = s->width;
319     YUV*p1 = &pp1[by*linex*16+bx*16];
320     YUV*p2 = &pp2[by*linex*16+bx*16];
321     int diffy=0, diffuv = 0;
322     int x,y;
323     for(y=0;y<16;y++) {
324 	for(x=0;x<16;x++) {
325 	    YUV*m = &p1[x];
326 	    YUV*n = &p2[x];
327 	    int y = m->y - n->y;
328 	    int u = m->u - n->u;
329 	    int v = m->v - n->v;
330 	    diffy += abs(y);
331 	    diffuv += abs(u)+abs(v);
332 	}
333 	p1+=linex;
334 	p2+=linex;
335     }
336     return diffy + diffuv/4;
337 }
338 
compare_pic_block(VIDEOSTREAM * s,block_t * b,YUV * pic,int bx,int by)339 static int compare_pic_block(VIDEOSTREAM*s, block_t* b, YUV*pic, int bx, int by)
340 {
341     int linex = s->width;
342     YUV*y1 = &pic[(by*2)*linex*8+bx*16];
343     YUV*y2 = &pic[(by*2)*linex*8+bx*16+8];
344     YUV*y3 = &pic[(by*2+1)*linex*8+bx*16];
345     YUV*y4 = &pic[(by*2+1)*linex*8+bx*16+8];
346     YUV*uv1 = y1;
347     YUV*uv2 = &y1[linex];
348     int diffy=0, diffuv = 0;
349     int x,y;
350     for(y=0;y<8;y++) {
351 	for(x=0;x<8;x++) {
352 	    int yy,u1,v1,u2,v2,u3,v3,u4,v4;
353 	    int y8x = y*8+x;
354 	    yy = y1[x].y - b->y1[y8x];
355 	    diffy += abs(yy);
356 	    yy = y2[x].y - b->y2[y8x];
357 	    diffy += abs(yy);
358 	    yy = y3[x].y - b->y3[y8x];
359 	    diffy += abs(yy);
360 	    yy = y4[x].y - b->y4[y8x];
361 	    diffy += abs(yy);
362 	    u1 = uv1[x*2].u - b->u[y8x];
363 	    v1 = uv1[x*2].v - b->v[y8x];
364 	    u2 = uv1[x*2+1].u - b->u[y8x];
365 	    v2 = uv1[x*2+1].v - b->v[y8x];
366 	    u3 = uv2[x*2].u - b->u[y8x];
367 	    v3 = uv2[x*2].v - b->v[y8x];
368 	    u4 = uv2[x*2+1].u - b->u[y8x];
369 	    v4 = uv2[x*2+1].v - b->v[y8x];
370 	    diffuv += (abs(u1)+abs(v1));
371 	    diffuv += (abs(u2)+abs(v2));
372 	    diffuv += (abs(u3)+abs(v3));
373 	    diffuv += (abs(u4)+abs(v4));
374 	}
375 	y1+=linex;
376 	y2+=linex;
377 	y3+=linex;
378 	y4+=linex;
379 	uv1+=linex*2;
380 	uv2+=linex*2;
381     }
382     return diffy + diffuv/4;
383 }
384 
valtodc(int val)385 static inline int valtodc(int val)
386 {
387     assert(val>=0);
388 
389     /* table 12/h.263 */
390 
391     //val+=4; //round
392     val/=8;
393     /* TODO: what to do for zero values? skip the block? */
394     if(val==0)
395 	return 1;
396     if(val==128)
397 	return 255;
398     if(val>254)
399 	return 254;
400     return val;
401 }
dctoval(int dc)402 static int dctoval(int dc)
403 {
404     int val;
405     assert(dc>0);
406     assert(dc!=128);
407     assert(dc<256);
408     /* table 12/h.263 */
409     val = dc*8;
410     if(val == 255*8)
411 	val = 128*8;
412     return val;
413 }
414 
415 /* TODO: we could also just let the caller pass only the string table[index] here */
codehuffman(TAG * tag,struct huffcode * table,int index)416 static int codehuffman(TAG*tag, struct huffcode*table, int index)
417 {
418     /* TODO: !optimize! */
419     int i=0;
420     while(table[index].code[i]) {
421 	if(table[index].code[i]=='0')
422 	    swf_SetBits(tag, 0, 1);
423 	else
424 	    swf_SetBits(tag, 1, 1);
425 	i++;
426     }
427     return i;
428 }
429 
quantize8x8(int * src,int * dest,int has_dc,int quant)430 static void quantize8x8(int*src, int*dest, int has_dc, int quant)
431 {
432     int t,pos=0;
433     double q = 1.0/(quant*2);
434     if(has_dc) {
435 	dest[0] = valtodc((int)src[0]); /*DC*/
436 	pos++;
437     }
438     for(t=pos;t<64;t++)
439     {
440 	//dest[t] = (int)src[t];
441     /* exact: if(quant&1){dest[t] = (dest[t]/quant - 1)/2;}else{dest[t] = ((dest[t]+1)/quant - 1)/2;} */
442         //if(quant&1){dest[t] = (dest[t]/quant - 1)/2;}else{dest[t] = ((dest[t]+1)/quant - 1)/2;}
443 	//dest[t] = dest[t]/(quant*2);
444 	dest[t] = (int)(src[t]*q);
445 	/* TODO: warn if this happens- the video will be buggy */
446 	if(dest[t]>127) dest[t]=127;
447 	if(dest[t]<-127) dest[t]=-127;
448     }
449 }
450 
dequantize8x8(int * b,int has_dc,int quant)451 static void dequantize8x8(int*b, int has_dc, int quant)
452 {
453     int t,pos=0;
454     if(has_dc) {
455 	b[0] = dctoval(b[0]); //DC
456 	pos++;
457     }
458     for(t=pos;t<64;t++) {
459 	if(b[t]) {
460 	    int sign = 0;
461 	    if(b[t]<0) {
462 		b[t] = -b[t];
463 		sign = 1;
464 	    }
465 
466 	    if(quant&1) {
467 		b[t] = quant*(2*b[t]+1); //-7,8,24,40
468 	    } else {
469 		b[t] = quant*(2*b[t]+1)-1; //-8,7,23,39
470 	    }
471 
472 	    if(sign)
473 		b[t] = -b[t];
474 	}
475 
476 	/* paragraph 6.2.2, "clipping of reconstruction levels": */
477 	if(b[t]>2047) b[t]=2047;
478 	if(b[t]<-2048) b[t]=-2048;
479     }
480 }
481 
hascoef(int * b,int has_dc)482 static int hascoef(int*b, int has_dc)
483 {
484     int t;
485     int pos=0;
486     if(has_dc)
487 	pos++;
488     for(t=pos;t<64;t++) {
489 	if(b[t])
490 	    return 1;
491     }
492     return 0;
493 }
494 
coefbits8x8(int * bb,int has_dc)495 static int coefbits8x8(int*bb, int has_dc)
496 {
497     int t;
498     int pos=0;
499     int bits=0;
500     int last;
501 
502     if(has_dc) {
503 	bits+=8;
504 	pos++;
505     }
506     for(last=63;last>=pos;last--) {
507 	if(bb[last])
508 	    break;
509     }
510     if(last < pos)
511 	return bits;
512     while(1) {
513 	int run=0, level=0, islast=0,t;
514 	while(!bb[pos] && pos<last) {
515 	    pos++;
516 	    run++;
517 	}
518 	if(pos==last)
519 	    islast=1;
520 	level=bb[pos];
521 	if(level<0) level=-level;
522 	assert(level);
523 	for(t=0;t<RLE_ESCAPE;t++) {
524 	    if(rle_params[t].run == run &&
525 	       rle_params[t].level == level &&
526 	       rle_params[t].last == islast) {
527 		bits += rle[t].len + 1;
528 		break;
529 	    }
530 	}
531 	if(t==RLE_ESCAPE) {
532 	    bits += rle[RLE_ESCAPE].len + 1 + 6 + 8;
533 	}
534 	if(islast)
535 	    break;
536 	pos++;
537     }
538     return bits;
539 }
540 
encode8x8(TAG * tag,int * bb,int has_dc,int has_tcoef)541 static int encode8x8(TAG*tag, int*bb, int has_dc, int has_tcoef)
542 {
543     int t;
544     int pos=0;
545     int bits=0;
546 
547     if(has_dc) {
548 	swf_SetBits(tag, bb[0], 8);
549 	bits += 8;
550 	pos++;
551     }
552 
553     if(has_tcoef) {
554 	int last;
555 	/* determine last non-null coefficient */
556 	for(last=63;last>=pos;last--) {
557 	    /* TODO: we could leave out small coefficients
558 	             after a certain point (32?) */
559 	    if(bb[last])
560 		break;
561 	}
562 	/* blocks without coefficients should not be included
563 	   in the cbpy/cbpc patterns: */
564 	assert(bb[last]);
565 
566 	while(1) {
567 	    int run=0;
568 	    int level=0;
569 	    int islast=0;
570 	    int sign=0;
571 	    int t;
572 	    while(!bb[pos] && pos<last) {
573 		pos++;
574 		run++;
575 	    }
576 	    if(pos==last)
577 		islast=1;
578 	    level=bb[pos];
579 	    assert(level);
580 	    if(level<0) {
581 		level = -level;
582 		sign = 1;
583 	    }
584 	    for(t=0;t<RLE_ESCAPE;t++) {
585 		/* TODO: lookup table */
586 		if(rle_params[t].run == run &&
587 		   rle_params[t].level == level &&
588 		   rle_params[t].last == islast) {
589 		    bits += codehuffman(tag, rle, t);
590 		    swf_SetBits(tag, sign, 1);
591 		    bits += 1;
592 		    break;
593 		}
594 	    }
595 	    if(t==RLE_ESCAPE) {
596 		bits += codehuffman(tag, rle, RLE_ESCAPE);
597 		level=bb[pos];
598 		/* table 14/h.263 */
599 		if(!level || level<-127 || level>127) {
600 		    fprintf(stderr, "Warning: Overflow- Level %d at pos %d\n", level, pos);
601 		    if(level<-127) level=-127;
602 		    if(level>127) level=127;
603 		}
604 
605 		assert(level);
606 		assert(level>=-127);
607 		assert(level<=127); //TODO: known to fail for pos=0 (with custom frames?)
608 
609 		swf_SetBits(tag, islast, 1);
610 		swf_SetBits(tag, run, 6);
611 		swf_SetBits(tag, level, 8); //FIXME: fixme??
612 		bits += 1 + 6 + 8;
613 	    }
614 
615 	    if(islast)
616 		break;
617 	    pos++;
618 	}
619     }
620     return bits;
621 }
622 
quantize(block_t * fb,block_t * b,int has_dc,int quant)623 static void quantize(block_t*fb, block_t*b, int has_dc, int quant)
624 {
625     quantize8x8(fb->y1, b->y1, has_dc, quant);
626     quantize8x8(fb->y2, b->y2, has_dc, quant);
627     quantize8x8(fb->y3, b->y3, has_dc, quant);
628     quantize8x8(fb->y4, b->y4, has_dc, quant);
629     quantize8x8(fb->u, b->u, has_dc, quant);
630     quantize8x8(fb->v, b->v, has_dc, quant);
631 }
632 
dodct(block_t * fb)633 static void dodct(block_t*fb)
634 {
635     dct(fb->y1); dct(fb->y2); dct(fb->y3); dct(fb->y4);
636     dct(fb->u);  dct(fb->v);
637     zigzag(fb->y1);
638     zigzag(fb->y2);
639     zigzag(fb->y3);
640     zigzag(fb->y4);
641     zigzag(fb->u);
642     zigzag(fb->v);
643 }
644 
dodctandquant(block_t * fb,block_t * b,int has_dc,int quant)645 static void dodctandquant(block_t*fb, block_t*b, int has_dc, int quant)
646 {
647     int t;
648     if(has_dc) {
649 	dodct(fb);
650 	quantize(fb,b,has_dc,quant);
651 	return;
652     }
653     preparequant(quant);
654     dct2(fb->y1,b->y1); dct2(fb->y2,b->y2); dct2(fb->y3,b->y3); dct2(fb->y4,b->y4);
655     dct2(fb->u,b->u);  dct2(fb->v,b->v);
656 
657     for(t=0;t<64;t++) {
658 	/* prepare for encoding (only values in (-127..-1,1..127) are
659 	   allowed as non-zero, non-dc values */
660 	if(b->y1[t]<-127) b->y1[t]=-127;
661 	if(b->y2[t]<-127) b->y2[t]=-127;
662 	if(b->y3[t]<-127) b->y3[t]=-127;
663 	if(b->y4[t]<-127) b->y4[t]=-127;
664 	if(b->u[t]<-127) b->u[t]=-127;
665 	if(b->v[t]<-127) b->v[t]=-127;
666 
667 	if(b->y1[t]>127) b->y1[t]=127;
668 	if(b->y2[t]>127) b->y2[t]=127;
669 	if(b->y3[t]>127) b->y3[t]=127;
670 	if(b->y4[t]>127) b->y4[t]=127;
671 	if(b->u[t]>127) b->u[t]=127;
672 	if(b->v[t]>127) b->v[t]=127;
673     }
674 }
675 
doidct(block_t * b)676 static void doidct(block_t*b)
677 {
678     block_t fb;
679     int t;
680     for(t=0;t<64;t++) {
681 	fb.y1[t] = b->y1[zigzagtable[t]];
682 	fb.y2[t] = b->y2[zigzagtable[t]];
683 	fb.y3[t] = b->y3[zigzagtable[t]];
684 	fb.y4[t] = b->y4[zigzagtable[t]];
685 	fb.u[t] = b->u[zigzagtable[t]];
686 	fb.v[t] = b->v[zigzagtable[t]];
687     }
688     idct(fb.y1); idct(fb.y2); idct(fb.y3); idct(fb.y4);
689     idct(fb.u);  idct(fb.v);
690 
691     memcpy(b, &fb, sizeof(block_t));
692 }
693 
truncateblock(block_t * b)694 static void truncateblock(block_t*b)
695 {
696     int t;
697     for(t=0;t<64;t++) {
698 	b->y1[t] = truncate256(b->y1[t]);
699 	b->y2[t] = truncate256(b->y2[t]);
700 	b->y3[t] = truncate256(b->y3[t]);
701 	b->y4[t] = truncate256(b->y4[t]);
702 	b->u[t] = truncate256(b->u[t]);
703 	b->v[t] = truncate256(b->v[t]);
704     }
705 }
706 
dequantize(block_t * b,int has_dc,int quant)707 static void dequantize(block_t*b, int has_dc, int quant)
708 {
709     dequantize8x8(b->y1, has_dc, quant);
710     dequantize8x8(b->y2, has_dc, quant);
711     dequantize8x8(b->y3, has_dc, quant);
712     dequantize8x8(b->y4, has_dc, quant);
713     dequantize8x8(b->u, has_dc, quant);
714     dequantize8x8(b->v, has_dc, quant);
715 }
716 
getblockpatterns(block_t * b,int * cbpybits,int * cbpcbits,int has_dc)717 static void getblockpatterns(block_t*b, int*cbpybits,int*cbpcbits, int has_dc)
718 {
719     *cbpybits = 0;
720     *cbpcbits = 0;
721 
722     *cbpybits|=hascoef(b->y1, has_dc)*8;
723     *cbpybits|=hascoef(b->y2, has_dc)*4;
724     *cbpybits|=hascoef(b->y3, has_dc)*2;
725     *cbpybits|=hascoef(b->y4, has_dc)*1;
726 
727     *cbpcbits|=hascoef(b->u, has_dc)*2;
728     *cbpcbits|=hascoef(b->v, has_dc)*1;
729 }
730 
setQuant(TAG * tag,int dquant)731 static void setQuant(TAG*tag, int dquant)
732 {
733     int code = 0;
734     /* 00 01 10 11
735        -1 -2 +1 +2
736     */
737     if(dquant == -1) {
738 	swf_SetBits(tag, 0x0, 2);
739     } else if(dquant == -2) {
740 	swf_SetBits(tag, 0x1, 2);
741     } else if(dquant == +1) {
742 	swf_SetBits(tag, 0x2, 2);
743     } else if(dquant == +2) {
744 	swf_SetBits(tag, 0x3, 2);
745     } else {
746 	assert(0*strlen("invalid dquant"));
747     }
748 }
749 
change_quant(int quant,int * dquant)750 static void change_quant(int quant, int*dquant)
751 {
752     /* TODO */
753     *dquant = 0;
754 }
755 
yuvdiff(block_t * a,block_t * b)756 static void yuvdiff(block_t*a, block_t*b)
757 {
758     int t;
759     for(t=0;t<64;t++) {
760 	a->y1[t] = (a->y1[t] - b->y1[t]);
761 	a->y2[t] = (a->y2[t] - b->y2[t]);
762 	a->y3[t] = (a->y3[t] - b->y3[t]);
763 	a->y4[t] = (a->y4[t] - b->y4[t]);
764 	a->u[t]  = (a->u[t] - b->u[t]);
765 	a->v[t]  = (a->v[t] - b->v[t]);
766     }
767 }
768 
predictmvd(VIDEOSTREAM * s,int bx,int by,int * px,int * py)769 static void predictmvd(VIDEOSTREAM*s, int bx, int by, int*px, int*py)
770 {
771     int i1,i2;
772     int x1,y1,x2,y2,x3,y3;
773     int x4,y4,p;
774     if(bx) {x1=s->mvdx[by*s->bbx+bx-1];
775 	    y1=s->mvdy[by*s->bbx+bx-1];
776     } else {x1=y1=0;}
777 
778     if(by) {x2=s->mvdx[(by-1)*s->bbx+bx];
779 	    y2=s->mvdy[(by-1)*s->bbx+bx];
780 	    if(bx<s->bbx-1) {
781 		x3=s->mvdx[(by-1)*s->bbx+bx+1];
782 		y3=s->mvdy[(by-1)*s->bbx+bx+1];
783 	    } else {
784 		x3=y3=0;
785 	    }
786            }
787     else   {x2=x3=x1;y2=y3=y1;}
788 
789 	   if((x1 <= x2 && x2 <= x3) ||
790 	      (x3 <= x2 && x2 <= x1)) {
791 	x4=x2;
792     } else if((x2 <= x1 && x1 <= x3) ||
793 	      (x3 <= x1 && x1 <= x2)) {
794 	x4=x1;
795     } else if((x1 <= x3 && x3 <= x2) ||
796 	      (x2 <= x3 && x3 <= x1)) {
797 	x4=x3;
798     } else {
799 	x4=0;
800 	assert(x4);
801     }
802 
803 	   if((y1 <= y2 && y2 <= y3) ||
804 	      (y3 <= y2 && y2 <= y1)) {
805 	y4=y2;
806     } else if((y2 <= y1 && y1 <= y3) ||
807 	      (y3 <= y1 && y1 <= y2)) {
808 	y4=y1;
809     } else if((y1 <= y3 && y3 <= y2) ||
810 	      (y2 <= y3 && y3 <= y1)) {
811 	y4=y3;
812     } else {
813 	y4=0;
814 	assert(y4);
815     }
816 
817     *px = x4;
818     *py = y4;
819     assert((x4>=-32 && x4<=31) && (y4>=-32 && y4<=31));
820 }
821 
mvd2index(int px,int py,int x,int y,int xy)822 static inline int mvd2index(int px, int py, int x, int y, int xy)
823 {
824 
825     if((x<-32 && x>31) || (y<-32 && y>31))
826 	fprintf(stderr, "(%d,%d)\n", x,y);
827     assert((x>=-32 && x<=31) && (y>=-32 && y<=31));
828     //assert((x&1)==0 && (y&1)==0);//for now
829     //assert((x&2)==0 && (y&2)==0);//for now(2)
830 
831     x-=px;
832     y-=py;
833 
834     if(xy)
835 	x=y;
836     x+=32;
837 
838     /* (x&63) */
839     if(x>63)
840 	x-=64;
841     if(x<0)
842 	x+=64;
843 
844     assert(x>=0 && x<64);
845     return x;
846 }
847 
848 typedef struct _iblockdata_t
849 {
850     block_t b; //transformed quantized coefficients
851     block_t reconstruction;
852     int bits;
853     int bx,by;
854     struct huffcode*ctable; //table to use for chrominance encoding (different for i-frames)
855     int iframe; // 1 if this is part of an iframe
856 } iblockdata_t;
857 
858 typedef struct _mvdblockdata_t
859 {
860     block_t b;
861     block_t fbold;
862     block_t reconstruction;
863     int xindex;
864     int yindex;
865     int movex;
866     int movey;
867     int bits;
868     int bx,by;
869 } mvdblockdata_t;
870 
prepareIBlock(VIDEOSTREAM * s,iblockdata_t * data,int bx,int by,block_t * fb,int * bits,int iframe)871 void prepareIBlock(VIDEOSTREAM*s, iblockdata_t*data, int bx, int by, block_t* fb, int*bits, int iframe)
872 {
873     /* consider I-block */
874     block_t fb_i;
875     block_t b;
876     int y,c;
877     struct huffcode*ctable;
878 
879     data->bx = bx;
880     data->by = by;
881 
882     data->iframe = iframe;
883     if(!iframe) {
884 	data->ctable = &mcbpc_inter[3*4];
885     } else {
886 	data->ctable = &mcbpc_intra[0];
887     }
888 
889     memcpy(&fb_i, fb, sizeof(block_t));
890     dodctandquant(&fb_i, &data->b, 1, s->quant);
891     getblockpatterns(&data->b, &y, &c, 1);
892     *bits = 0;
893     if(!data->iframe) {
894 	*bits += 1; //cod
895     }
896     *bits += data->ctable[c].len;
897     *bits += cbpy[y].len;
898     *bits += coefbits8x8(data->b.y1, 1);
899     *bits += coefbits8x8(data->b.y2, 1);
900     *bits += coefbits8x8(data->b.y3, 1);
901     *bits += coefbits8x8(data->b.y4, 1);
902     *bits += coefbits8x8(data->b.u, 1);
903     *bits += coefbits8x8(data->b.v, 1);
904     data->bits = *bits;
905 
906     /* -- reconstruction -- */
907     memcpy(&data->reconstruction,&data->b,sizeof(block_t));
908     dequantize(&data->reconstruction, 1, s->quant);
909     doidct(&data->reconstruction);
910     truncateblock(&data->reconstruction);
911 }
912 
writeIBlock(VIDEOSTREAM * s,TAG * tag,iblockdata_t * data)913 int writeIBlock(VIDEOSTREAM*s, TAG*tag, iblockdata_t*data)
914 {
915     int c = 0, y = 0;
916     int has_dc=1;
917     int bits = 0;
918     block_t b;
919 
920     getblockpatterns(&data->b, &y, &c, has_dc);
921     if(!data->iframe) {
922 	swf_SetBits(tag,0,1); bits += 1; // COD
923     }
924     bits += codehuffman(tag, data->ctable, c);
925     bits += codehuffman(tag, cbpy, y);
926 
927     /* luminance */
928     bits += encode8x8(tag, data->b.y1, has_dc, y&8);
929     bits += encode8x8(tag, data->b.y2, has_dc, y&4);
930     bits += encode8x8(tag, data->b.y3, has_dc, y&2);
931     bits += encode8x8(tag, data->b.y4, has_dc, y&1);
932 
933     /* chrominance */
934     bits += encode8x8(tag, data->b.u, has_dc, c&2);
935     bits += encode8x8(tag, data->b.v, has_dc, c&1);
936 
937     copy_block_pic(s, s->current, &data->reconstruction, data->bx, data->by);
938     assert(data->bits == bits);
939     return bits;
940 }
941 
getmvdbits(VIDEOSTREAM * s,block_t * fb,int bx,int by,int hx,int hy)942 int getmvdbits(VIDEOSTREAM*s,block_t*fb, int bx,int by,int hx,int hy)
943 {
944     block_t b;
945     block_t fbold;
946     block_t fbdiff;
947     int bits = 0;
948     memcpy(&fbdiff, fb, sizeof(block_t));
949     getmvdregion(&fbold, s->oldpic, bx, by, hx, hy, s->linex);
950     yuvdiff(&fbdiff, &fbold);
951     dodctandquant(&fbdiff, &b, 0, s->quant);
952     bits += coefbits8x8(b.y1, 0);
953     bits += coefbits8x8(b.y2, 0);
954     bits += coefbits8x8(b.y3, 0);
955     bits += coefbits8x8(b.y4, 0);
956     bits += coefbits8x8(b.u, 0);
957     bits += coefbits8x8(b.v, 0);
958     return bits;
959 }
960 
prepareMVDBlock(VIDEOSTREAM * s,mvdblockdata_t * data,int bx,int by,block_t * fb,int * bits)961 void prepareMVDBlock(VIDEOSTREAM*s, mvdblockdata_t*data, int bx, int by, block_t* fb, int*bits)
962 { /* consider mvd(x,y)-block */
963 
964     int t;
965     int y,c;
966     block_t fbdiff;
967     int predictmvdx;
968     int predictmvdy;
969 
970     data->bx = bx;
971     data->by = by;
972     predictmvd(s,bx,by,&predictmvdx,&predictmvdy);
973 
974     data->bits = 65535;
975     data->movex=0;
976     data->movey=0;
977 
978     if(s->do_motion) {
979 	int hx,hy;
980 	int bestx=0,besty=0,bestbits=65536;
981 	int startx=-32,endx=31;
982 	int starty=-32,endy=31;
983 
984 	if(!bx) startx=0;
985 	if(!by) starty=0;
986 	if(bx==s->bbx-1) endx=0;
987 	if(by==s->bby-1) endy=0;
988 
989 	for(hx=startx;hx<=endx;hx+=4)
990 	for(hy=starty;hy<=endy;hy+=4)
991 	{
992 	    int bits = 0;
993 	    bits = getmvdbits(s,fb,bx,by,hx,hy);
994 	    if(bits<bestbits) {
995 		bestbits = bits;
996 		bestx = hx;
997 		besty = hy;
998 	    }
999 	}
1000 
1001 	if(bestx-3 > startx) startx = bestx-3;
1002 	if(besty-3 > starty) starty = besty-3;
1003 	if(bestx+3 < endx) endx = bestx+3;
1004 	if(besty+3 < endy) endy = besty+3;
1005 
1006 	for(hx=startx;hx<=endx;hx++)
1007 	for(hy=starty;hy<=endy;hy++)
1008 	{
1009 	    int bits = 0;
1010 	    bits = getmvdbits(s,fb,bx,by,hx,hy);
1011 	    if(bits<bestbits) {
1012 		bestbits = bits;
1013 		bestx = hx;
1014 		besty = hy;
1015 	    }
1016 	}
1017 	data->movex = bestx;
1018 	data->movey = besty;
1019     }
1020 
1021     memcpy(&fbdiff, fb, sizeof(block_t));
1022     getmvdregion(&data->fbold, s->oldpic, bx, by, data->movex, data->movey, s->linex);
1023     yuvdiff(&fbdiff, &data->fbold);
1024     dodctandquant(&fbdiff, &data->b, 0, s->quant);
1025     getblockpatterns(&data->b, &y, &c, 0);
1026 
1027     data->xindex = mvd2index(predictmvdx, predictmvdy, data->movex, data->movey, 0);
1028     data->yindex = mvd2index(predictmvdx, predictmvdy, data->movex, data->movey, 1);
1029 
1030     *bits = 1; //cod
1031     *bits += mcbpc_inter[0*4+c].len;
1032     *bits += cbpy[y^15].len;
1033     *bits += mvd[data->xindex].len; // (0,0)
1034     *bits += mvd[data->yindex].len;
1035     *bits += coefbits8x8(data->b.y1, 0);
1036     *bits += coefbits8x8(data->b.y2, 0);
1037     *bits += coefbits8x8(data->b.y3, 0);
1038     *bits += coefbits8x8(data->b.y4, 0);
1039     *bits += coefbits8x8(data->b.u, 0);
1040     *bits += coefbits8x8(data->b.v, 0);
1041     data->bits = *bits;
1042 
1043     /* -- reconstruction -- */
1044     memcpy(&data->reconstruction, &data->b, sizeof(block_t));
1045     dequantize(&data->reconstruction, 0, s->quant);
1046     doidct(&data->reconstruction);
1047     for(t=0;t<64;t++) {
1048 	data->reconstruction.y1[t] =
1049 	    truncate256(data->reconstruction.y1[t] + (int)data->fbold.y1[t]);
1050 	data->reconstruction.y2[t] =
1051 	    truncate256(data->reconstruction.y2[t] + (int)data->fbold.y2[t]);
1052 	data->reconstruction.y3[t] =
1053 	    truncate256(data->reconstruction.y3[t] + (int)data->fbold.y3[t]);
1054 	data->reconstruction.y4[t] =
1055 	    truncate256(data->reconstruction.y4[t] + (int)data->fbold.y4[t]);
1056 	data->reconstruction.u[t] =
1057 	    truncate256(data->reconstruction.u[t] + (int)data->fbold.u[t]);
1058 	data->reconstruction.v[t] =
1059 	    truncate256(data->reconstruction.v[t] + (int)data->fbold.v[t]);
1060     }
1061 }
1062 
writeMVDBlock(VIDEOSTREAM * s,TAG * tag,mvdblockdata_t * data)1063 int writeMVDBlock(VIDEOSTREAM*s, TAG*tag, mvdblockdata_t*data)
1064 {
1065     int c = 0, y = 0;
1066     int t;
1067     int has_dc=0; // mvd w/o mvd24
1068     /* mvd (0,0) block (mode=0) */
1069     int mode = 0;
1070     int bx = data->bx;
1071     int by = data->by;
1072     int bits = 0;
1073 
1074     getblockpatterns(&data->b, &y, &c, has_dc);
1075     swf_SetBits(tag,0,1); bits += 1; // COD
1076     bits += codehuffman(tag, mcbpc_inter, mode*4+c);
1077     bits += codehuffman(tag, cbpy, y^15);
1078 
1079     /* vector */
1080     bits += codehuffman(tag, mvd, data->xindex);
1081     bits += codehuffman(tag, mvd, data->yindex);
1082 
1083     /* luminance */
1084     bits += encode8x8(tag, data->b.y1, has_dc, y&8);
1085     bits += encode8x8(tag, data->b.y2, has_dc, y&4);
1086     bits += encode8x8(tag, data->b.y3, has_dc, y&2);
1087     bits += encode8x8(tag, data->b.y4, has_dc, y&1);
1088 
1089     /* chrominance */
1090     bits += encode8x8(tag, data->b.u, has_dc, c&2);
1091     bits += encode8x8(tag, data->b.v, has_dc, c&1);
1092 
1093     s->mvdx[by*s->bbx+bx] = data->movex;
1094     s->mvdy[by*s->bbx+bx] = data->movey;
1095 
1096     copy_block_pic(s, s->current, &data->reconstruction, data->bx, data->by);
1097     assert(data->bits == bits);
1098     return bits;
1099 }
1100 
encode_PFrame_block(TAG * tag,VIDEOSTREAM * s,int bx,int by)1101 static int encode_PFrame_block(TAG*tag, VIDEOSTREAM*s, int bx, int by)
1102 {
1103     block_t fb;
1104     int diff1,diff2;
1105     int bits_i;
1106     int bits_vxy;
1107 
1108     iblockdata_t iblock;
1109     mvdblockdata_t mvdblock;
1110 
1111     getregion(&fb, s->current, bx, by, s->linex);
1112     prepareIBlock(s, &iblock, bx, by, &fb, &bits_i, 0);
1113 
1114     /* encoded last frame <=> original current block: */
1115     diff1 = compare_pic_pic(s, s->current, s->oldpic, bx, by);
1116     /* encoded current frame <=> original current block: */
1117     diff2 = compare_pic_block(s, &iblock.reconstruction, s->current, bx, by);
1118 
1119     if(diff1 <= diff2) {
1120 	swf_SetBits(tag, 1,1); /* cod=1, block skipped */
1121 	/* copy the region from the last frame so that we have a complete reconstruction */
1122 	copyregion(s, s->current, s->oldpic, bx, by);
1123 	return 1;
1124     }
1125     prepareMVDBlock(s, &mvdblock, bx, by, &fb, &bits_vxy);
1126 
1127     if(bits_i > bits_vxy) {
1128 	return writeMVDBlock(s, tag, &mvdblock);
1129     } else {
1130 	return writeIBlock(s, tag, &iblock);
1131     }
1132 }
1133 
1134 /* should be called encode_IFrameBlock */
encode_IFrame_block(TAG * tag,VIDEOSTREAM * s,int bx,int by)1135 static void encode_IFrame_block(TAG*tag, VIDEOSTREAM*s, int bx, int by)
1136 {
1137     block_t fb;
1138     iblockdata_t data;
1139     int bits;
1140 
1141     getregion(&fb, s->current, bx, by, s->width);
1142     prepareIBlock(s, &data, bx, by, &fb, &bits, 1);
1143     writeIBlock(s, tag, &data);
1144 }
1145 
1146 #ifdef MAIN
1147 static int bmid = 0;
1148 
setdbgpic(TAG * tag,RGBA * pic,int width,int height)1149 void setdbgpic(TAG*tag, RGBA*pic, int width, int height)
1150 {
1151     MATRIX m;
1152     tag = tag->prev;
1153 
1154     tag = swf_InsertTag(tag,ST_REMOVEOBJECT2);
1155     swf_SetU16(tag, 133);
1156 
1157     tag = swf_InsertTag(tag, ST_DEFINEBITSLOSSLESS);
1158     swf_SetU16(tag, 1000+bmid);
1159     swf_SetLosslessBits(tag, width, height, (void*)pic, BMF_32BIT);
1160 
1161     tag = swf_InsertTag(tag, ST_DEFINESHAPE);
1162     swf_SetU16(tag, 2000+bmid);
1163     swf_ShapeSetBitmapRect(tag, 1000+bmid, width, height);
1164 
1165     tag = swf_InsertTag(tag,ST_PLACEOBJECT2);
1166     swf_GetMatrix(0,&m);
1167     m.tx = width*20;
1168     swf_ObjectPlace(tag, 2000+bmid, 133, &m, 0, 0);
1169 
1170     bmid++;
1171 }
1172 #endif
1173 
1174 #define TYPE_IFRAME 0
1175 #define TYPE_PFRAME 1
1176 
writeHeader(TAG * tag,int width,int height,int frame,int quant,int type)1177 static void writeHeader(TAG*tag, int width, int height, int frame, int quant, int type)
1178 {
1179     U32 i32;
1180     swf_SetU16(tag, frame);
1181     swf_SetBits(tag, 1, 17); /* picture start code*/
1182     swf_SetBits(tag, 0, 5); /* version=0, version 1 would optimize rle behaviour*/
1183     swf_SetBits(tag, frame, 8); /* time reference */
1184 
1185     /* write dimensions, taking advantage of some predefined sizes
1186        if the opportunity presents itself */
1187     i32 = width<<16|height;
1188     switch(i32)
1189     {
1190 	case 352<<16|288: swf_SetBits(tag, 2, 3);break;
1191 	case 176<<16|144: swf_SetBits(tag, 3, 3);break;
1192 	case 128<<16|96: swf_SetBits(tag, 4, 3);break;
1193 	case 320<<16|240: swf_SetBits(tag, 5, 3);break;
1194 	case 160<<16|120: swf_SetBits(tag, 6, 3);break;
1195 	default:
1196 	    if(width>255 || height>255) {
1197 		swf_SetBits(tag, 1, 3);
1198 		swf_SetBits(tag, width, 16);
1199 		swf_SetBits(tag, height, 16);
1200 	    } else {
1201 		swf_SetBits(tag, 0, 3);
1202 		swf_SetBits(tag, width, 8);
1203 		swf_SetBits(tag, height, 8);
1204 	    }
1205     }
1206 
1207     swf_SetBits(tag, type, 2); /* I-Frame or P-Frame */
1208     swf_SetBits(tag, 0, 1); /* No deblock filter */
1209     assert(quant>0);
1210     assert(quant<32);
1211     swf_SetBits(tag, quant, 5); /* quantizer (1-31), may be updated later on*/
1212     swf_SetBits(tag, 0, 1); /* No extra info */
1213 }
1214 
swf_SetVideoStreamIFrame(TAG * tag,VIDEOSTREAM * s,RGBA * pic,int quant)1215 void swf_SetVideoStreamIFrame(TAG*tag, VIDEOSTREAM*s, RGBA*pic, int quant)
1216 {
1217     int bx, by;
1218 
1219     if(quant<1) quant=1;
1220     if(quant>31) quant=31;
1221     s->quant = quant;
1222 
1223     writeHeader(tag, s->width, s->height, s->frame, quant, TYPE_IFRAME);
1224 
1225     /* fixme: should fill with 0,128,128, not 0,0,0 */
1226     memset(s->current, 0, s->linex*s->height*sizeof(YUV));
1227 
1228     rgb2yuv(s->current, pic, s->linex, s->olinex, s->owidth, s->oheight);
1229 
1230     for(by=0;by<s->bby;by++)
1231     {
1232 	for(bx=0;bx<s->bbx;bx++)
1233 	{
1234 	    encode_IFrame_block(tag, s, bx, by);
1235 	}
1236     }
1237     s->frame++;
1238     memcpy(s->oldpic, s->current, s->width*s->height*sizeof(YUV));
1239 }
swf_SetVideoStreamBlackFrame(TAG * tag,VIDEOSTREAM * s)1240 void swf_SetVideoStreamBlackFrame(TAG*tag, VIDEOSTREAM*s)
1241 {
1242     int bx, by;
1243     int quant = 31;
1244     int x,y;
1245     s->quant = quant;
1246 
1247     writeHeader(tag, s->width, s->height, s->frame, quant, TYPE_IFRAME);
1248 
1249     for(y=0;y<s->height;y++)
1250     for(x=0;x<s->width;x++) {
1251 	s->current[y*s->width+x].y = 0;
1252 	s->current[y*s->width+x].u = 128;
1253 	s->current[y*s->width+x].v = 128;
1254     }
1255     for(x=0;x<16;x++)
1256     for(y=0;y<16;y++) {
1257 	s->current[y*s->width+x].y = 64;
1258 	s->current[y*s->width+x].u = 128;
1259 	s->current[y*s->width+x].v = 128;
1260     }
1261 
1262     for(by=0;by<s->bby;by++)
1263     {
1264 	for(bx=0;bx<s->bbx;bx++)
1265 	{
1266 	    encode_IFrame_block(tag, s, bx, by);
1267 	}
1268     }
1269     s->frame++;
1270     memcpy(s->oldpic, s->current, s->width*s->height*sizeof(YUV));
1271 }
1272 
swf_SetVideoStreamPFrame(TAG * tag,VIDEOSTREAM * s,RGBA * pic,int quant)1273 void swf_SetVideoStreamPFrame(TAG*tag, VIDEOSTREAM*s, RGBA*pic, int quant)
1274 {
1275     int bx, by;
1276 
1277     if(quant<1) quant=1;
1278     if(quant>31) quant=31;
1279     s->quant = quant;
1280 
1281     writeHeader(tag, s->width, s->height, s->frame, quant, TYPE_PFRAME);
1282 
1283     /* fixme: should fill with 0,128,128, not 0,0,0 */
1284     memset(s->current, 0, s->linex*s->height*sizeof(YUV));
1285 
1286     rgb2yuv(s->current, pic, s->linex, s->olinex, s->owidth, s->oheight);
1287     memset(s->mvdx, 0, s->bbx*s->bby*sizeof(int));
1288     memset(s->mvdy, 0, s->bbx*s->bby*sizeof(int));
1289 
1290     for(by=0;by<s->bby;by++)
1291     {
1292 	for(bx=0;bx<s->bbx;bx++)
1293 	{
1294 	    encode_PFrame_block(tag, s, bx, by);
1295 	}
1296     }
1297     s->frame++;
1298     memcpy(s->oldpic, s->current, s->width*s->height*sizeof(YUV));
1299 
1300 #ifdef MAIN
1301 #ifdef PNG
1302     yuv2rgb(pic, s->current, s->linex, s->width, s->height);
1303     setdbgpic(tag, pic, s->width, s->height);
1304 #endif
1305 #endif
1306 }
1307 
swf_SetVideoStreamMover(TAG * tag,VIDEOSTREAM * s,signed char * movex,signed char * movey,void ** pictures,int quant)1308 void swf_SetVideoStreamMover(TAG*tag, VIDEOSTREAM*s, signed char* movex, signed char* movey, void**pictures, int quant)
1309 {
1310     int bx, by;
1311     YUV pic[16*16];
1312 
1313     if(quant<1) quant=1;
1314     if(quant>31) quant=31;
1315     s->quant = quant;
1316 
1317     writeHeader(tag, s->width, s->height, s->frame, quant, TYPE_PFRAME);
1318 
1319     memset(s->mvdx, 0, s->bbx*s->bby*sizeof(int));
1320     memset(s->mvdy, 0, s->bbx*s->bby*sizeof(int));
1321 
1322     for(by=0;by<s->bby;by++)
1323     {
1324 	for(bx=0;bx<s->bbx;bx++)
1325 	{
1326 	    int predictmvdx=0, predictmvdy=0;
1327 	    int mvx=movex[by*s->bbx+bx];
1328 	    int mvy=movey[by*s->bbx+bx];
1329 	    void*picture = pictures?pictures[by*s->bbx+bx]:0;
1330 
1331 	    if(mvx<-32) mvx=-32;
1332 	    if(mvx>31) mvx=31;
1333 	    if(mvy<-32) mvy=-32;
1334 	    if(mvy>31) mvy=31;
1335 
1336 	    if(mvx == 0 && mvy == 0 && picture == 0) {
1337 		swf_SetBits(tag,1,1); // COD skip
1338 	    } else {
1339 		int mode = 0;
1340 		int has_dc=0;
1341 		int y=0,c=0;
1342 		block_t b;
1343 		block_t b2;
1344 
1345 		swf_SetBits(tag,0,1); // COD
1346 
1347 		if(mvx==0 && mvy==0 && picture) { // only picture
1348 		    mode = 3;
1349 		    has_dc = 1;
1350 		}
1351 
1352 		if(picture) {
1353 		    RGBA* picblock = (RGBA*)picture;
1354 		    rgb2yuv(pic, picblock,16,16,16,16);
1355 		    /* TODO: if has_dc!=1, subtract 128 from rgb values */
1356 		    getregion(&b, pic, 0,0,16);
1357 		    dodctandquant(&b, &b2, 1, s->quant);
1358 		    getblockpatterns(&b2, &y, &c, 1);
1359 		} else {
1360 		    y=0;c=0;
1361 		}
1362 
1363 		codehuffman(tag, mcbpc_inter, mode*4+c);
1364 		codehuffman(tag, cbpy, mode==3?y:y^15);
1365 
1366 		if(mode < 3) {
1367 		    /* has motion vector */
1368 		    predictmvd(s,bx,by,&predictmvdx,&predictmvdy);
1369 		    codehuffman(tag, mvd, mvd2index(predictmvdx, predictmvdy, mvx, mvy, 0));
1370 		    codehuffman(tag, mvd, mvd2index(predictmvdx, predictmvdy, mvx, mvy, 1));
1371 		    s->mvdx[by*s->bbx+bx] = mvx;
1372 		    s->mvdy[by*s->bbx+bx] = mvy;
1373 		}
1374 
1375 		if(has_dc||y||c) {
1376 		    encode8x8(tag, b2.y1, has_dc, y&8);
1377 		    encode8x8(tag, b2.y2, has_dc, y&4);
1378 		    encode8x8(tag, b2.y3, has_dc, y&2);
1379 		    encode8x8(tag, b2.y4, has_dc, y&1);
1380 		    encode8x8(tag, b2.u, has_dc, c&2);
1381 		    encode8x8(tag, b2.v, has_dc, c&1);
1382 		}
1383 	    }
1384 	}
1385     }
1386     s->frame++;
1387 }
1388 
1389 #define TESTS
1390 #ifdef TESTS
test_copy_diff()1391 void test_copy_diff()
1392 {
1393     VIDEOSTREAM stream;
1394     VIDEOSTREAM* s = &stream;
1395     TAG*tag;
1396     RGBA*pic = (RGBA*)rfx_alloc(256*256*sizeof(RGBA));
1397     block_t fb;
1398     int x,y;
1399     int bx,by;
1400     for(x=0;x<256;x++)
1401     for(y=0;y<256;y++) {
1402 	pic[y*256+x].r = x*y;
1403 	pic[y*256+x].g = x+y;
1404 	pic[y*256+x].b = (x+1)%(y+1);
1405     }
1406     tag = swf_InsertTag(0, ST_DEFINEVIDEOSTREAM);
1407     swf_SetU16(tag, 33);
1408     swf_SetVideoStreamDefine(tag, s, 10, 256, 256);
1409 
1410     rgb2yuv(s->current, pic, s->linex, s->olinex, s->owidth, s->oheight);
1411     for(by=0;by<16;by++)
1412     for(bx=0;bx<16;bx++) {
1413 	int diff1,diff2;
1414 	/* test1: does compare pic pic return zero for identical blocks? */
1415 	diff1 = compare_pic_pic(s, s->current, s->current, bx, by);
1416 	assert(!diff1);
1417 	/* test2: do blocks which are copied back return zero diff? */
1418 	getregion(&fb, s->current, bx, by, s->linex);
1419 	copy_block_pic(s, s->oldpic, &fb, bx, by);
1420 	diff1 = compare_pic_block(s, &fb, s->oldpic, bx, by);
1421 	assert(!diff1);
1422 	/* test3: does compare_pic_block return the same result as compare_pic_pic? */
1423 	getregion(&fb, s->current, 15-bx, 15-by, s->linex);
1424 	copy_block_pic(s, s->oldpic, &fb, bx, by);
1425 	diff1 = compare_pic_block(s, &fb, s->current, bx, by);
1426 	diff2 = compare_pic_pic(s, s->current, s->oldpic, bx, by);
1427 	assert(diff1 == diff2);
1428     }
1429 }
1430 
1431 #endif
1432 
1433 #ifdef MAIN
1434 #include "png.h"
1435 
compileSWFActionCode(const char * script,int version,void ** data,int * len)1436 int compileSWFActionCode(const char *script, int version, void**data, int*len) {return 0;}
1437 
mkblack()1438 void mkblack()
1439 {
1440     SWF swf;
1441     SWFPLACEOBJECT obj;
1442     int frames = 88;
1443     int width = 160;
1444     int height = 112;
1445     int x,y;
1446     TAG*tag = 0;
1447     RGBA rgb;
1448     RGBA* pic = 0;
1449     VIDEOSTREAM stream;
1450 
1451     pic = rfx_calloc(width*height*4);
1452 
1453     memset(&swf,0,sizeof(SWF));
1454     memset(&obj,0,sizeof(obj));
1455 
1456     swf.fileVersion    = 6;
1457     swf.frameRate      = 15*256;
1458     swf.movieSize.xmax = 20*width;
1459     swf.movieSize.ymax = 20*height;
1460 
1461     swf.firstTag = swf_InsertTag(NULL,ST_SETBACKGROUNDCOLOR);
1462     tag = swf.firstTag;
1463     rgb.r = 0x00;rgb.g = 0x30;rgb.b = 0xff;
1464     swf_SetRGB(tag,&rgb);
1465 
1466     tag = swf_InsertTag(tag, ST_DEFINEVIDEOSTREAM);
1467     swf_SetU16(tag, 1);
1468     swf_SetVideoStreamDefine(tag, &stream, frames, width, height);
1469     stream.do_motion = 0;
1470 
1471     for(y=0;y<height;y++)  {
1472 	for(x=0;x<width;x++) {
1473 	    int dx = x/16;
1474 	    int dy = y/16;
1475 
1476 	    pic[y*width+x].r = 0;
1477 	    pic[y*width+x].g = 0;
1478 	    pic[y*width+x].b = 0;
1479 	    pic[y*width+x].a = 0;
1480 	}
1481     }
1482     tag = swf_InsertTag(tag, ST_VIDEOFRAME);
1483     swf_SetU16(tag, 1);
1484 
1485     swf_SetVideoStreamIFrame(tag, &stream, pic, 7);
1486 
1487     tag = swf_InsertTag(tag, ST_PLACEOBJECT2);
1488     swf_GetPlaceObject(0, &obj);
1489 
1490     obj.depth = 4;
1491     obj.id = 1;
1492 
1493     swf_SetPlaceObject(tag,&obj);
1494 
1495     tag = swf_InsertTag(tag, ST_SHOWFRAME);
1496 
1497     swf_VideoStreamClear(&stream);
1498 
1499     tag = swf_InsertTag(tag, ST_END);
1500 
1501     int fi = open("black.swf", O_WRONLY|O_CREAT|O_TRUNC, 0644);
1502     if(swf_WriteSWF(fi,&swf)<0) {
1503 	fprintf(stderr,"WriteSWF() failed.\n");
1504     }
1505     close(fi);
1506     swf_FreeTags(&swf);
1507 }
1508 
main(int argn,char * argv[])1509 int main(int argn, char*argv[])
1510 {
1511     int fi;
1512     int t;
1513     SWF swf;
1514     TAG * tag;
1515     RGBA* pic, *pic2, rgb;
1516     SWFPLACEOBJECT obj;
1517     unsigned width = 0;
1518     unsigned height = 0;
1519     int frames = 10;
1520     int framerate = 29;
1521     unsigned char*data;
1522     char* fname = "/home/kramm/pics/peppers_fromjpg.png";
1523     //char* fname = "/home/kramm/pics/baboon.png";
1524     VIDEOSTREAM stream;
1525     double d = 1.0;
1526 
1527 #ifdef TESTS
1528     test_copy_diff();
1529 #endif
1530 
1531     mkblack();
1532 
1533     memset(&stream, 0, sizeof(stream));
1534 
1535     png_load(fname, &width, &height, &data);
1536     pic = (RGBA*)rfx_alloc(width*height*sizeof(RGBA));
1537     pic2 = (RGBA*)rfx_alloc(width*height*sizeof(RGBA));
1538     memcpy(pic, data, width*height*sizeof(RGBA));
1539     rfx_free(data);
1540 
1541     printf("Compressing %s, size %dx%d\n", fname, width, height);
1542 
1543     memset(&swf,0,sizeof(SWF));
1544     memset(&obj,0,sizeof(obj));
1545 
1546     swf.fileVersion    = 6;
1547     swf.frameRate      = framerate*256;
1548     swf.movieSize.xmax = 20*width*2;
1549     swf.movieSize.ymax = 20*height;
1550 
1551     swf.firstTag = swf_InsertTag(NULL,ST_SETBACKGROUNDCOLOR);
1552     tag = swf.firstTag;
1553     rgb.r = 0x00;rgb.g = 0x30;rgb.b = 0xff;
1554     swf_SetRGB(tag,&rgb);
1555 
1556     tag = swf_InsertTag(tag, ST_DEFINEVIDEOSTREAM);
1557     swf_SetU16(tag, 33);
1558     swf_SetVideoStreamDefine(tag, &stream, frames, width, height);
1559     stream.do_motion = 0;
1560 
1561     //srand48(time(0));
1562 
1563     for(t=0;t<frames;t++)
1564     {
1565 	int x,y;
1566 	double xx,yy;
1567 	for(y=0,yy=0;y<height;y++,yy+=d)  {
1568 	    RGBA*line = &pic[((int)yy)*width];
1569 	    for(x=0,xx=0;x<width;x++,xx+=d) {
1570 		int dx = x/16;
1571 		int dy = y/16;
1572 		if(dx==0 && dy==0) {
1573 		    pic2[y*width+x] = line[((int)xx)];
1574 		    pic2[y*width+x].r+=2;
1575 		    pic2[y*width+x].g+=2;
1576 		    pic2[y*width+x].b+=2;
1577 		} else {
1578 		    //pic2[y*width+x] = line[((int)xx)];
1579 		    //pic2[y*width+x].r = lrand48();//line[((int)xx)];
1580 		    //pic2[y*width+x].g = lrand48();//line[((int)xx)];
1581 		    //pic2[y*width+x].b = lrand48();//line[((int)xx)];
1582 		    pic2[y*width+x].r = 0;
1583 		    pic2[y*width+x].g = 0;
1584 		    pic2[y*width+x].b = 0;
1585 		}
1586 		/*if(dx==16 && dy==16)
1587 		    pic2[y*width+x] = pic[(y-16*16)*width+(x-16*16)];*/
1588 		/*if(dx<=0 && dy<=0) {
1589 		    pic2[y*width+x] = line[((int)xx)];*/
1590 		/*if(x==0 && y==0) {
1591 		    RGBA color;
1592 		    memset(&color, 0, sizeof(RGBA));
1593 		    pic2[y*width+x] = color;*/
1594 		/*} else  {
1595 		    RGBA color;
1596 		    color.r = lrand48();
1597 		    color.g = lrand48();
1598 		    color.b = lrand48();
1599 		    color.a = 0;
1600 		    pic2[y*width+x] = color;
1601 		}*/
1602 	    }
1603 	}
1604 	printf("frame:%d\n", t);fflush(stdout);
1605 
1606 	if(t==1)
1607 	    break;
1608 
1609 	tag = swf_InsertTag(tag, ST_VIDEOFRAME);
1610 	swf_SetU16(tag, 33);
1611 	if(t==0)
1612 	    swf_SetVideoStreamIFrame(tag, &stream, pic2, 7);
1613 	else {
1614 	    swf_SetVideoStreamPFrame(tag, &stream, pic2, 7);
1615 	}
1616 
1617 	tag = swf_InsertTag(tag, ST_PLACEOBJECT2);
1618 	swf_GetPlaceObject(0, &obj);
1619 	if(t==0) {
1620 	    obj.depth = 1;
1621 	    obj.id = 33;
1622 	} else {
1623 	    obj.move = 1;
1624 	    obj.depth = 1;
1625 	    obj.ratio = t;
1626 	}
1627 	swf_SetPlaceObject(tag,&obj);
1628 
1629 	tag = swf_InsertTag(tag, ST_SHOWFRAME);
1630 	d-=0.015;
1631     }
1632     swf_VideoStreamClear(&stream);
1633 
1634     tag = swf_InsertTag(tag, ST_END);
1635 
1636     fi = open("video3.swf", O_WRONLY|O_CREAT|O_TRUNC, 0644);
1637     if(swf_WriteSWF(fi,&swf)<0) {
1638 	fprintf(stderr,"WriteSWF() failed.\n");
1639     }
1640     close(fi);
1641     swf_FreeTags(&swf);
1642     return 0;
1643 }
1644 #undef MAIN
1645 #endif
1646