2 * Copyright (C) 2004-2010 Michael Niedermayer <michaelni@gmx.at>
4 * This file is part of FFmpeg.
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 #include "libavutil/attributes.h"
25 void ff_slice_buffer_init(slice_buffer * buf, int line_count, int max_allocated_lines, int line_width, IDWTELEM * base_buffer)
29 buf->base_buffer = base_buffer;
30 buf->line_count = line_count;
31 buf->line_width = line_width;
32 buf->data_count = max_allocated_lines;
33 buf->line = av_mallocz (sizeof(IDWTELEM *) * line_count);
34 buf->data_stack = av_malloc (sizeof(IDWTELEM *) * max_allocated_lines);
36 for(i = 0; i < max_allocated_lines; i++){
37 buf->data_stack[i] = av_malloc (sizeof(IDWTELEM) * line_width);
40 buf->data_stack_top = max_allocated_lines - 1;
43 IDWTELEM * ff_slice_buffer_load_line(slice_buffer * buf, int line)
47 assert(buf->data_stack_top >= 0);
48 // assert(!buf->line[line]);
50 return buf->line[line];
52 buffer = buf->data_stack[buf->data_stack_top];
53 buf->data_stack_top--;
54 buf->line[line] = buffer;
59 void ff_slice_buffer_release(slice_buffer * buf, int line)
63 assert(line >= 0 && line < buf->line_count);
64 assert(buf->line[line]);
66 buffer = buf->line[line];
67 buf->data_stack_top++;
68 buf->data_stack[buf->data_stack_top] = buffer;
69 buf->line[line] = NULL;
72 void ff_slice_buffer_flush(slice_buffer * buf)
75 for(i = 0; i < buf->line_count; i++){
77 ff_slice_buffer_release(buf, i);
81 void ff_slice_buffer_destroy(slice_buffer * buf)
84 ff_slice_buffer_flush(buf);
86 for(i = buf->data_count - 1; i >= 0; i--){
87 av_freep(&buf->data_stack[i]);
89 av_freep(&buf->data_stack);
93 static inline int mirror(int v, int m){
94 while((unsigned)v > (unsigned)m){
101 static av_always_inline void
102 lift(DWTELEM *dst, DWTELEM *src, DWTELEM *ref,
103 int dst_step, int src_step, int ref_step,
104 int width, int mul, int add, int shift,
105 int highpass, int inverse){
106 const int mirror_left= !highpass;
107 const int mirror_right= (width&1) ^ highpass;
108 const int w= (width>>1) - 1 + (highpass & width);
111 #define LIFT(src, ref, inv) ((src) + ((inv) ? - (ref) : + (ref)))
113 dst[0] = LIFT(src[0], ((mul*2*ref[0]+add)>>shift), inverse);
120 LIFT(src[i*src_step],
121 ((mul*(ref[i*ref_step] + ref[(i+1)*ref_step])+add)>>shift),
127 LIFT(src[w*src_step],
128 ((mul*2*ref[w*ref_step]+add)>>shift),
133 static av_always_inline void
134 inv_lift(IDWTELEM *dst, IDWTELEM *src, IDWTELEM *ref,
135 int dst_step, int src_step, int ref_step,
136 int width, int mul, int add, int shift,
137 int highpass, int inverse){
138 const int mirror_left= !highpass;
139 const int mirror_right= (width&1) ^ highpass;
140 const int w= (width>>1) - 1 + (highpass & width);
143 #define LIFT(src, ref, inv) ((src) + ((inv) ? - (ref) : + (ref)))
145 dst[0] = LIFT(src[0], ((mul*2*ref[0]+add)>>shift), inverse);
152 LIFT(src[i*src_step],
153 ((mul*(ref[i*ref_step] + ref[(i+1)*ref_step])+add)>>shift),
159 LIFT(src[w*src_step],
160 ((mul*2*ref[w*ref_step]+add)>>shift),
166 static av_always_inline void
167 liftS(DWTELEM *dst, DWTELEM *src, DWTELEM *ref,
168 int dst_step, int src_step, int ref_step,
169 int width, int mul, int add, int shift,
170 int highpass, int inverse){
171 const int mirror_left= !highpass;
172 const int mirror_right= (width&1) ^ highpass;
173 const int w= (width>>1) - 1 + (highpass & width);
177 #define LIFTS(src, ref, inv) \
179 (src) + (((ref) + 4*(src))>>shift): \
180 -((-16*(src) + (ref) + add/4 + 1 + (5<<25))/(5*4) - (1<<23)))
182 dst[0] = LIFTS(src[0], mul*2*ref[0]+add, inverse);
189 LIFTS(src[i*src_step],
190 mul*(ref[i*ref_step] + ref[(i+1)*ref_step])+add,
196 LIFTS(src[w*src_step], mul*2*ref[w*ref_step]+add, inverse);
199 static av_always_inline void
200 inv_liftS(IDWTELEM *dst, IDWTELEM *src, IDWTELEM *ref,
201 int dst_step, int src_step, int ref_step,
202 int width, int mul, int add, int shift,
203 int highpass, int inverse){
204 const int mirror_left= !highpass;
205 const int mirror_right= (width&1) ^ highpass;
206 const int w= (width>>1) - 1 + (highpass & width);
210 #define LIFTS(src, ref, inv) \
212 (src) + (((ref) + 4*(src))>>shift): \
213 -((-16*(src) + (ref) + add/4 + 1 + (5<<25))/(5*4) - (1<<23)))
215 dst[0] = LIFTS(src[0], mul*2*ref[0]+add, inverse);
222 LIFTS(src[i*src_step],
223 mul*(ref[i*ref_step] + ref[(i+1)*ref_step])+add,
229 LIFTS(src[w*src_step], mul*2*ref[w*ref_step]+add, inverse);
234 static void horizontal_decompose53i(DWTELEM *b, int width){
236 const int width2= width>>1;
238 const int w2= (width+1)>>1;
240 for(x=0; x<width2; x++){
242 temp[x+w2]= b[2*x + 1];
256 for(x=1; x+1<width2; x+=2){
260 A2 += (A1 + A3 + 2)>>2;
264 A1= temp[x+1+width2];
267 A4 += (A1 + A3 + 2)>>2;
273 A2 += (A1 + A3 + 2)>>2;
278 lift(b+w2, temp+w2, temp, 1, 1, 1, width, -1, 0, 1, 1, 0);
279 lift(b , temp , b+w2, 1, 1, 1, width, 1, 2, 2, 0, 0);
283 static void vertical_decompose53iH0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
286 for(i=0; i<width; i++){
287 b1[i] -= (b0[i] + b2[i])>>1;
291 static void vertical_decompose53iL0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
294 for(i=0; i<width; i++){
295 b1[i] += (b0[i] + b2[i] + 2)>>2;
299 static void spatial_decompose53i(DWTELEM *buffer, int width, int height, int stride){
301 DWTELEM *b0= buffer + mirror(-2-1, height-1)*stride;
302 DWTELEM *b1= buffer + mirror(-2 , height-1)*stride;
304 for(y=-2; y<height; y+=2){
305 DWTELEM *b2= buffer + mirror(y+1, height-1)*stride;
306 DWTELEM *b3= buffer + mirror(y+2, height-1)*stride;
308 if(y+1<(unsigned)height) horizontal_decompose53i(b2, width);
309 if(y+2<(unsigned)height) horizontal_decompose53i(b3, width);
311 if(y+1<(unsigned)height) vertical_decompose53iH0(b1, b2, b3, width);
312 if(y+0<(unsigned)height) vertical_decompose53iL0(b0, b1, b2, width);
319 static void horizontal_decompose97i(DWTELEM *b, int width){
321 const int w2= (width+1)>>1;
323 lift (temp+w2, b +1, b , 1, 2, 2, width, W_AM, W_AO, W_AS, 1, 1);
324 liftS(temp , b , temp+w2, 1, 2, 1, width, W_BM, W_BO, W_BS, 0, 0);
325 lift (b +w2, temp+w2, temp , 1, 1, 1, width, W_CM, W_CO, W_CS, 1, 0);
326 lift (b , temp , b +w2, 1, 1, 1, width, W_DM, W_DO, W_DS, 0, 0);
330 static void vertical_decompose97iH0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
333 for(i=0; i<width; i++){
334 b1[i] -= (W_AM*(b0[i] + b2[i])+W_AO)>>W_AS;
338 static void vertical_decompose97iH1(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
341 for(i=0; i<width; i++){
342 b1[i] += (W_CM*(b0[i] + b2[i])+W_CO)>>W_CS;
346 static void vertical_decompose97iL0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
349 for(i=0; i<width; i++){
351 b1[i] -= (W_BM*(b0[i] + b2[i])+W_BO)>>W_BS;
353 b1[i] = (16*4*b1[i] - 4*(b0[i] + b2[i]) + W_BO*5 + (5<<27)) / (5*16) - (1<<23);
358 static void vertical_decompose97iL1(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
361 for(i=0; i<width; i++){
362 b1[i] += (W_DM*(b0[i] + b2[i])+W_DO)>>W_DS;
366 static void spatial_decompose97i(DWTELEM *buffer, int width, int height, int stride){
368 DWTELEM *b0= buffer + mirror(-4-1, height-1)*stride;
369 DWTELEM *b1= buffer + mirror(-4 , height-1)*stride;
370 DWTELEM *b2= buffer + mirror(-4+1, height-1)*stride;
371 DWTELEM *b3= buffer + mirror(-4+2, height-1)*stride;
373 for(y=-4; y<height; y+=2){
374 DWTELEM *b4= buffer + mirror(y+3, height-1)*stride;
375 DWTELEM *b5= buffer + mirror(y+4, height-1)*stride;
377 if(y+3<(unsigned)height) horizontal_decompose97i(b4, width);
378 if(y+4<(unsigned)height) horizontal_decompose97i(b5, width);
380 if(y+3<(unsigned)height) vertical_decompose97iH0(b3, b4, b5, width);
381 if(y+2<(unsigned)height) vertical_decompose97iL0(b2, b3, b4, width);
382 if(y+1<(unsigned)height) vertical_decompose97iH1(b1, b2, b3, width);
383 if(y+0<(unsigned)height) vertical_decompose97iL1(b0, b1, b2, width);
392 void ff_spatial_dwt(DWTELEM *buffer, int width, int height, int stride, int type, int decomposition_count){
395 for(level=0; level<decomposition_count; level++){
397 case DWT_97: spatial_decompose97i(buffer, width>>level, height>>level, stride<<level); break;
398 case DWT_53: spatial_decompose53i(buffer, width>>level, height>>level, stride<<level); break;
403 static void horizontal_compose53i(IDWTELEM *b, int width){
404 IDWTELEM temp[width];
405 const int width2= width>>1;
406 const int w2= (width+1)>>1;
409 for(x=0; x<width2; x++){
411 temp[2*x + 1]= b[x+w2];
416 b[0] = temp[0] - ((temp[1]+1)>>1);
417 for(x=2; x<width-1; x+=2){
418 b[x ] = temp[x ] - ((temp[x-1] + temp[x+1]+2)>>2);
419 b[x-1] = temp[x-1] + ((b [x-2] + b [x ]+1)>>1);
422 b[x ] = temp[x ] - ((temp[x-1]+1)>>1);
423 b[x-1] = temp[x-1] + ((b [x-2] + b [x ]+1)>>1);
425 b[x-1] = temp[x-1] + b[x-2];
428 static void vertical_compose53iH0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width){
431 for(i=0; i<width; i++){
432 b1[i] += (b0[i] + b2[i])>>1;
436 static void vertical_compose53iL0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width){
439 for(i=0; i<width; i++){
440 b1[i] -= (b0[i] + b2[i] + 2)>>2;
444 static void spatial_compose53i_buffered_init(DWTCompose *cs, slice_buffer * sb, int height, int stride_line){
445 cs->b0 = slice_buffer_get_line(sb, mirror(-1-1, height-1) * stride_line);
446 cs->b1 = slice_buffer_get_line(sb, mirror(-1 , height-1) * stride_line);
450 static void spatial_compose53i_init(DWTCompose *cs, IDWTELEM *buffer, int height, int stride){
451 cs->b0 = buffer + mirror(-1-1, height-1)*stride;
452 cs->b1 = buffer + mirror(-1 , height-1)*stride;
456 static void spatial_compose53i_dy_buffered(DWTCompose *cs, slice_buffer * sb, int width, int height, int stride_line){
459 IDWTELEM *b0= cs->b0;
460 IDWTELEM *b1= cs->b1;
461 IDWTELEM *b2= slice_buffer_get_line(sb, mirror(y+1, height-1) * stride_line);
462 IDWTELEM *b3= slice_buffer_get_line(sb, mirror(y+2, height-1) * stride_line);
464 if(y+1<(unsigned)height && y<(unsigned)height){
467 for(x=0; x<width; x++){
468 b2[x] -= (b1[x] + b3[x] + 2)>>2;
469 b1[x] += (b0[x] + b2[x])>>1;
472 if(y+1<(unsigned)height) vertical_compose53iL0(b1, b2, b3, width);
473 if(y+0<(unsigned)height) vertical_compose53iH0(b0, b1, b2, width);
476 if(y-1<(unsigned)height) horizontal_compose53i(b0, width);
477 if(y+0<(unsigned)height) horizontal_compose53i(b1, width);
484 static void spatial_compose53i_dy(DWTCompose *cs, IDWTELEM *buffer, int width, int height, int stride){
486 IDWTELEM *b0= cs->b0;
487 IDWTELEM *b1= cs->b1;
488 IDWTELEM *b2= buffer + mirror(y+1, height-1)*stride;
489 IDWTELEM *b3= buffer + mirror(y+2, height-1)*stride;
491 if(y+1<(unsigned)height) vertical_compose53iL0(b1, b2, b3, width);
492 if(y+0<(unsigned)height) vertical_compose53iH0(b0, b1, b2, width);
494 if(y-1<(unsigned)height) horizontal_compose53i(b0, width);
495 if(y+0<(unsigned)height) horizontal_compose53i(b1, width);
502 static void av_unused spatial_compose53i(IDWTELEM *buffer, int width, int height, int stride){
504 spatial_compose53i_init(&cs, buffer, height, stride);
505 while(cs.y <= height)
506 spatial_compose53i_dy(&cs, buffer, width, height, stride);
510 void ff_snow_horizontal_compose97i(IDWTELEM *b, int width){
511 IDWTELEM temp[width];
512 const int w2= (width+1)>>1;
514 #if 0 //maybe more understadable but slower
515 inv_lift (temp , b , b +w2, 2, 1, 1, width, W_DM, W_DO, W_DS, 0, 1);
516 inv_lift (temp+1 , b +w2, temp , 2, 1, 2, width, W_CM, W_CO, W_CS, 1, 1);
518 inv_liftS(b , temp , temp+1 , 2, 2, 2, width, W_BM, W_BO, W_BS, 0, 1);
519 inv_lift (b+1 , temp+1 , b , 2, 2, 2, width, W_AM, W_AO, W_AS, 1, 0);
522 temp[0] = b[0] - ((3*b[w2]+2)>>2);
523 for(x=1; x<(width>>1); x++){
524 temp[2*x ] = b[x ] - ((3*(b [x+w2-1] + b[x+w2])+4)>>3);
525 temp[2*x-1] = b[x+w2-1] - temp[2*x-2] - temp[2*x];
528 temp[2*x ] = b[x ] - ((3*b [x+w2-1]+2)>>2);
529 temp[2*x-1] = b[x+w2-1] - temp[2*x-2] - temp[2*x];
531 temp[2*x-1] = b[x+w2-1] - 2*temp[2*x-2];
533 b[0] = temp[0] + ((2*temp[0] + temp[1]+4)>>3);
534 for(x=2; x<width-1; x+=2){
535 b[x ] = temp[x ] + ((4*temp[x ] + temp[x-1] + temp[x+1]+8)>>4);
536 b[x-1] = temp[x-1] + ((3*(b [x-2] + b [x ] ))>>1);
539 b[x ] = temp[x ] + ((2*temp[x ] + temp[x-1]+4)>>3);
540 b[x-1] = temp[x-1] + ((3*(b [x-2] + b [x ] ))>>1);
542 b[x-1] = temp[x-1] + 3*b [x-2];
546 static void vertical_compose97iH0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width){
549 for(i=0; i<width; i++){
550 b1[i] += (W_AM*(b0[i] + b2[i])+W_AO)>>W_AS;
554 static void vertical_compose97iH1(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width){
557 for(i=0; i<width; i++){
558 b1[i] -= (W_CM*(b0[i] + b2[i])+W_CO)>>W_CS;
562 static void vertical_compose97iL0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width){
565 for(i=0; i<width; i++){
567 b1[i] += (W_BM*(b0[i] + b2[i])+W_BO)>>W_BS;
569 b1[i] += (W_BM*(b0[i] + b2[i])+4*b1[i]+W_BO)>>W_BS;
574 static void vertical_compose97iL1(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width){
577 for(i=0; i<width; i++){
578 b1[i] -= (W_DM*(b0[i] + b2[i])+W_DO)>>W_DS;
582 void ff_snow_vertical_compose97i(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, IDWTELEM *b5, int width){
585 for(i=0; i<width; i++){
586 b4[i] -= (W_DM*(b3[i] + b5[i])+W_DO)>>W_DS;
587 b3[i] -= (W_CM*(b2[i] + b4[i])+W_CO)>>W_CS;
589 b2[i] += (W_BM*(b1[i] + b3[i])+W_BO)>>W_BS;
591 b2[i] += (W_BM*(b1[i] + b3[i])+4*b2[i]+W_BO)>>W_BS;
593 b1[i] += (W_AM*(b0[i] + b2[i])+W_AO)>>W_AS;
597 static void spatial_compose97i_buffered_init(DWTCompose *cs, slice_buffer * sb, int height, int stride_line){
598 cs->b0 = slice_buffer_get_line(sb, mirror(-3-1, height-1) * stride_line);
599 cs->b1 = slice_buffer_get_line(sb, mirror(-3 , height-1) * stride_line);
600 cs->b2 = slice_buffer_get_line(sb, mirror(-3+1, height-1) * stride_line);
601 cs->b3 = slice_buffer_get_line(sb, mirror(-3+2, height-1) * stride_line);
605 static void spatial_compose97i_init(DWTCompose *cs, IDWTELEM *buffer, int height, int stride){
606 cs->b0 = buffer + mirror(-3-1, height-1)*stride;
607 cs->b1 = buffer + mirror(-3 , height-1)*stride;
608 cs->b2 = buffer + mirror(-3+1, height-1)*stride;
609 cs->b3 = buffer + mirror(-3+2, height-1)*stride;
613 static void spatial_compose97i_dy_buffered(DWTContext *dsp, DWTCompose *cs, slice_buffer * sb, int width, int height, int stride_line){
616 IDWTELEM *b0= cs->b0;
617 IDWTELEM *b1= cs->b1;
618 IDWTELEM *b2= cs->b2;
619 IDWTELEM *b3= cs->b3;
620 IDWTELEM *b4= slice_buffer_get_line(sb, mirror(y + 3, height - 1) * stride_line);
621 IDWTELEM *b5= slice_buffer_get_line(sb, mirror(y + 4, height - 1) * stride_line);
623 if(y>0 && y+4<height){
624 dsp->vertical_compose97i(b0, b1, b2, b3, b4, b5, width);
626 if(y+3<(unsigned)height) vertical_compose97iL1(b3, b4, b5, width);
627 if(y+2<(unsigned)height) vertical_compose97iH1(b2, b3, b4, width);
628 if(y+1<(unsigned)height) vertical_compose97iL0(b1, b2, b3, width);
629 if(y+0<(unsigned)height) vertical_compose97iH0(b0, b1, b2, width);
632 if(y-1<(unsigned)height) dsp->horizontal_compose97i(b0, width);
633 if(y+0<(unsigned)height) dsp->horizontal_compose97i(b1, width);
642 static void spatial_compose97i_dy(DWTCompose *cs, IDWTELEM *buffer, int width, int height, int stride){
644 IDWTELEM *b0= cs->b0;
645 IDWTELEM *b1= cs->b1;
646 IDWTELEM *b2= cs->b2;
647 IDWTELEM *b3= cs->b3;
648 IDWTELEM *b4= buffer + mirror(y+3, height-1)*stride;
649 IDWTELEM *b5= buffer + mirror(y+4, height-1)*stride;
651 if(y+3<(unsigned)height) vertical_compose97iL1(b3, b4, b5, width);
652 if(y+2<(unsigned)height) vertical_compose97iH1(b2, b3, b4, width);
653 if(y+1<(unsigned)height) vertical_compose97iL0(b1, b2, b3, width);
654 if(y+0<(unsigned)height) vertical_compose97iH0(b0, b1, b2, width);
656 if(y-1<(unsigned)height) ff_snow_horizontal_compose97i(b0, width);
657 if(y+0<(unsigned)height) ff_snow_horizontal_compose97i(b1, width);
666 static void av_unused spatial_compose97i(IDWTELEM *buffer, int width, int height, int stride){
668 spatial_compose97i_init(&cs, buffer, height, stride);
669 while(cs.y <= height)
670 spatial_compose97i_dy(&cs, buffer, width, height, stride);
673 void ff_spatial_idwt_buffered_init(DWTCompose *cs, slice_buffer * sb, int width, int height, int stride_line, int type, int decomposition_count){
675 for(level=decomposition_count-1; level>=0; level--){
677 case DWT_97: spatial_compose97i_buffered_init(cs+level, sb, height>>level, stride_line<<level); break;
678 case DWT_53: spatial_compose53i_buffered_init(cs+level, sb, height>>level, stride_line<<level); break;
683 void ff_spatial_idwt_buffered_slice(DWTContext *dsp, DWTCompose *cs, slice_buffer * slice_buf, int width, int height, int stride_line, int type, int decomposition_count, int y){
684 const int support = type==1 ? 3 : 5;
688 for(level=decomposition_count-1; level>=0; level--){
689 while(cs[level].y <= FFMIN((y>>level)+support, height>>level)){
691 case DWT_97: spatial_compose97i_dy_buffered(dsp, cs+level, slice_buf, width>>level, height>>level, stride_line<<level);
693 case DWT_53: spatial_compose53i_dy_buffered(cs+level, slice_buf, width>>level, height>>level, stride_line<<level);
700 void ff_spatial_idwt_init(DWTCompose *cs, IDWTELEM *buffer, int width, int height, int stride, int type, int decomposition_count){
702 for(level=decomposition_count-1; level>=0; level--){
704 case DWT_97: spatial_compose97i_init(cs+level, buffer, height>>level, stride<<level); break;
705 case DWT_53: spatial_compose53i_init(cs+level, buffer, height>>level, stride<<level); break;
710 void ff_spatial_idwt_slice(DWTCompose *cs, IDWTELEM *buffer, int width, int height, int stride, int type, int decomposition_count, int y){
711 const int support = type==1 ? 3 : 5;
715 for(level=decomposition_count-1; level>=0; level--){
716 while(cs[level].y <= FFMIN((y>>level)+support, height>>level)){
718 case DWT_97: spatial_compose97i_dy(cs+level, buffer, width>>level, height>>level, stride<<level);
720 case DWT_53: spatial_compose53i_dy(cs+level, buffer, width>>level, height>>level, stride<<level);
727 void ff_spatial_idwt(IDWTELEM *buffer, int width, int height, int stride, int type, int decomposition_count){
728 DWTCompose cs[MAX_DECOMPOSITIONS];
730 ff_spatial_idwt_init(cs, buffer, width, height, stride, type, decomposition_count);
731 for(y=0; y<height; y+=4)
732 ff_spatial_idwt_slice(cs, buffer, width, height, stride, type, decomposition_count, y);
735 static inline int w_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int w, int h, int type){
737 const int dec_count= w==8 ? 3 : 4;
740 static const int scale[2][2][4][4]={
744 {268, 239, 239, 213},
748 // 9/7 16x16 or 32x32 dec=4
749 {344, 310, 310, 280},
757 {275, 245, 245, 218},
761 // 5/3 16x16 or 32x32 dec=4
762 {352, 317, 317, 286},
770 for (i = 0; i < h; i++) {
771 for (j = 0; j < w; j+=4) {
772 tmp[32*i+j+0] = (pix1[j+0] - pix2[j+0])<<4;
773 tmp[32*i+j+1] = (pix1[j+1] - pix2[j+1])<<4;
774 tmp[32*i+j+2] = (pix1[j+2] - pix2[j+2])<<4;
775 tmp[32*i+j+3] = (pix1[j+3] - pix2[j+3])<<4;
781 ff_spatial_dwt(tmp, w, h, 32, type, dec_count);
785 for(level=0; level<dec_count; level++){
786 for(ori= level ? 1 : 0; ori<4; ori++){
787 int size= w>>(dec_count-level);
788 int sx= (ori&1) ? size : 0;
789 int stride= 32<<(dec_count-level);
790 int sy= (ori&2) ? stride>>1 : 0;
792 for(i=0; i<size; i++){
793 for(j=0; j<size; j++){
794 int v= tmp[sx + sy + i*stride + j] * scale[type][dec_count-3][level][ori];
804 static int w53_8_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
805 return w_c(v, pix1, pix2, line_size, 8, h, 1);
808 static int w97_8_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
809 return w_c(v, pix1, pix2, line_size, 8, h, 0);
812 static int w53_16_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
813 return w_c(v, pix1, pix2, line_size, 16, h, 1);
816 static int w97_16_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
817 return w_c(v, pix1, pix2, line_size, 16, h, 0);
820 int ff_w53_32_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
821 return w_c(v, pix1, pix2, line_size, 32, h, 1);
824 int ff_w97_32_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
825 return w_c(v, pix1, pix2, line_size, 32, h, 0);
828 void ff_dsputil_init_dwt(DSPContext *c)
836 void ff_dwt_init(DWTContext *c)
838 c->vertical_compose97i = ff_snow_vertical_compose97i;
839 c->horizontal_compose97i = ff_snow_horizontal_compose97i;
840 c->inner_add_yblock = ff_snow_inner_add_yblock;
842 if (HAVE_MMX) ff_dwt_init_x86(c);