2 * Copyright (C) 2004-2010 Michael Niedermayer <michaelni@gmx.at>
3 * Copyright (C) 2008 David Conrad
5 * This file is part of FFmpeg.
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.
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.
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
22 #include "libavutil/attributes.h"
23 #include "libavutil/avassert.h"
24 #include "libavutil/common.h"
28 int ff_slice_buffer_init(slice_buffer *buf, int line_count,
29 int max_allocated_lines, int line_width,
30 IDWTELEM *base_buffer)
34 buf->base_buffer = base_buffer;
35 buf->line_count = line_count;
36 buf->line_width = line_width;
37 buf->data_count = max_allocated_lines;
38 buf->line = av_mallocz_array(line_count, sizeof(IDWTELEM *));
40 return AVERROR(ENOMEM);
41 buf->data_stack = av_malloc_array(max_allocated_lines, sizeof(IDWTELEM *));
42 if (!buf->data_stack) {
44 return AVERROR(ENOMEM);
47 for (i = 0; i < max_allocated_lines; i++) {
48 buf->data_stack[i] = av_malloc_array(line_width, sizeof(IDWTELEM));
49 if (!buf->data_stack[i]) {
51 av_freep(&buf->data_stack[i]);
52 av_freep(&buf->data_stack);
54 return AVERROR(ENOMEM);
58 buf->data_stack_top = max_allocated_lines - 1;
62 IDWTELEM *ff_slice_buffer_load_line(slice_buffer *buf, int line)
66 av_assert0(buf->data_stack_top >= 0);
67 // av_assert1(!buf->line[line]);
69 return buf->line[line];
71 buffer = buf->data_stack[buf->data_stack_top];
72 buf->data_stack_top--;
73 buf->line[line] = buffer;
78 void ff_slice_buffer_release(slice_buffer *buf, int line)
82 av_assert1(line >= 0 && line < buf->line_count);
83 av_assert1(buf->line[line]);
85 buffer = buf->line[line];
86 buf->data_stack_top++;
87 buf->data_stack[buf->data_stack_top] = buffer;
88 buf->line[line] = NULL;
91 void ff_slice_buffer_flush(slice_buffer *buf)
98 for (i = 0; i < buf->line_count; i++)
100 ff_slice_buffer_release(buf, i);
103 void ff_slice_buffer_destroy(slice_buffer *buf)
106 ff_slice_buffer_flush(buf);
109 for (i = buf->data_count - 1; i >= 0; i--)
110 av_freep(&buf->data_stack[i]);
111 av_freep(&buf->data_stack);
112 av_freep(&buf->line);
115 static inline int mirror(int v, int m)
117 while ((unsigned)v > (unsigned)m) {
125 static av_always_inline void lift(DWTELEM *dst, DWTELEM *src, DWTELEM *ref,
126 int dst_step, int src_step, int ref_step,
127 int width, int mul, int add, int shift,
128 int highpass, int inverse)
130 const int mirror_left = !highpass;
131 const int mirror_right = (width & 1) ^ highpass;
132 const int w = (width >> 1) - 1 + (highpass & width);
135 #define LIFT(src, ref, inv) ((src) + ((inv) ? -(ref) : +(ref)))
137 dst[0] = LIFT(src[0], ((mul * 2 * ref[0] + add) >> shift), inverse);
142 for (i = 0; i < w; i++)
143 dst[i * dst_step] = LIFT(src[i * src_step],
144 ((mul * (ref[i * ref_step] +
145 ref[(i + 1) * ref_step]) +
150 dst[w * dst_step] = LIFT(src[w * src_step],
151 ((mul * 2 * ref[w * ref_step] + add) >> shift),
155 static av_always_inline void liftS(DWTELEM *dst, DWTELEM *src, DWTELEM *ref,
156 int dst_step, int src_step, int ref_step,
157 int width, int mul, int add, int shift,
158 int highpass, int inverse)
160 const int mirror_left = !highpass;
161 const int mirror_right = (width & 1) ^ highpass;
162 const int w = (width >> 1) - 1 + (highpass & width);
165 av_assert1(shift == 4);
166 #define LIFTS(src, ref, inv) \
167 ((inv) ? (src) + (((ref) + 4 * (src)) >> shift) \
168 : -((-16 * (src) + (ref) + add / \
169 4 + 1 + (5 << 25)) / (5 * 4) - (1 << 23)))
171 dst[0] = LIFTS(src[0], mul * 2 * ref[0] + add, inverse);
176 for (i = 0; i < w; i++)
177 dst[i * dst_step] = LIFTS(src[i * src_step],
178 mul * (ref[i * ref_step] +
179 ref[(i + 1) * ref_step]) + add,
183 dst[w * dst_step] = LIFTS(src[w * src_step],
184 mul * 2 * ref[w * ref_step] + add,
188 static void horizontal_decompose53i(DWTELEM *b, DWTELEM *temp, int width)
190 const int width2 = width >> 1;
192 const int w2 = (width + 1) >> 1;
194 for (x = 0; x < width2; x++) {
196 temp[x + w2] = b[2 * x + 1];
200 lift(b + w2, temp + w2, temp, 1, 1, 1, width, -1, 0, 1, 1, 0);
201 lift(b, temp, b + w2, 1, 1, 1, width, 1, 2, 2, 0, 0);
204 static void vertical_decompose53iH0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2,
209 for (i = 0; i < width; i++)
210 b1[i] -= (b0[i] + b2[i]) >> 1;
213 static void vertical_decompose53iL0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2,
218 for (i = 0; i < width; i++)
219 b1[i] += (b0[i] + b2[i] + 2) >> 2;
222 static void spatial_decompose53i(DWTELEM *buffer, DWTELEM *temp,
223 int width, int height, int stride)
226 DWTELEM *b0 = buffer + mirror(-2 - 1, height - 1) * stride;
227 DWTELEM *b1 = buffer + mirror(-2, height - 1) * stride;
229 for (y = -2; y < height; y += 2) {
230 DWTELEM *b2 = buffer + mirror(y + 1, height - 1) * stride;
231 DWTELEM *b3 = buffer + mirror(y + 2, height - 1) * stride;
233 if (y + 1 < (unsigned)height)
234 horizontal_decompose53i(b2, temp, width);
235 if (y + 2 < (unsigned)height)
236 horizontal_decompose53i(b3, temp, width);
238 if (y + 1 < (unsigned)height)
239 vertical_decompose53iH0(b1, b2, b3, width);
240 if (y + 0 < (unsigned)height)
241 vertical_decompose53iL0(b0, b1, b2, width);
248 static void horizontal_decompose97i(DWTELEM *b, DWTELEM *temp, int width)
250 const int w2 = (width + 1) >> 1;
252 lift(temp + w2, b + 1, b, 1, 2, 2, width, W_AM, W_AO, W_AS, 1, 1);
253 liftS(temp, b, temp + w2, 1, 2, 1, width, W_BM, W_BO, W_BS, 0, 0);
254 lift(b + w2, temp + w2, temp, 1, 1, 1, width, W_CM, W_CO, W_CS, 1, 0);
255 lift(b, temp, b + w2, 1, 1, 1, width, W_DM, W_DO, W_DS, 0, 0);
258 static void vertical_decompose97iH0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2,
263 for (i = 0; i < width; i++)
264 b1[i] -= (W_AM * (b0[i] + b2[i]) + W_AO) >> W_AS;
267 static void vertical_decompose97iH1(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2,
272 for (i = 0; i < width; i++)
273 b1[i] += (W_CM * (b0[i] + b2[i]) + W_CO) >> W_CS;
276 static void vertical_decompose97iL0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2,
281 for (i = 0; i < width; i++)
282 b1[i] = (16 * 4 * b1[i] - 4 * (b0[i] + b2[i]) + W_BO * 5 + (5 << 27)) /
283 (5 * 16) - (1 << 23);
286 static void vertical_decompose97iL1(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2,
291 for (i = 0; i < width; i++)
292 b1[i] += (W_DM * (b0[i] + b2[i]) + W_DO) >> W_DS;
295 static void spatial_decompose97i(DWTELEM *buffer, DWTELEM *temp,
296 int width, int height, int stride)
299 DWTELEM *b0 = buffer + mirror(-4 - 1, height - 1) * stride;
300 DWTELEM *b1 = buffer + mirror(-4, height - 1) * stride;
301 DWTELEM *b2 = buffer + mirror(-4 + 1, height - 1) * stride;
302 DWTELEM *b3 = buffer + mirror(-4 + 2, height - 1) * stride;
304 for (y = -4; y < height; y += 2) {
305 DWTELEM *b4 = buffer + mirror(y + 3, height - 1) * stride;
306 DWTELEM *b5 = buffer + mirror(y + 4, height - 1) * stride;
308 if (y + 3 < (unsigned)height)
309 horizontal_decompose97i(b4, temp, width);
310 if (y + 4 < (unsigned)height)
311 horizontal_decompose97i(b5, temp, width);
313 if (y + 3 < (unsigned)height)
314 vertical_decompose97iH0(b3, b4, b5, width);
315 if (y + 2 < (unsigned)height)
316 vertical_decompose97iL0(b2, b3, b4, width);
317 if (y + 1 < (unsigned)height)
318 vertical_decompose97iH1(b1, b2, b3, width);
319 if (y + 0 < (unsigned)height)
320 vertical_decompose97iL1(b0, b1, b2, width);
329 void ff_spatial_dwt(DWTELEM *buffer, DWTELEM *temp, int width, int height,
330 int stride, int type, int decomposition_count)
334 for (level = 0; level < decomposition_count; level++) {
337 spatial_decompose97i(buffer, temp,
338 width >> level, height >> level,
342 spatial_decompose53i(buffer, temp,
343 width >> level, height >> level,
350 static void horizontal_compose53i(IDWTELEM *b, IDWTELEM *temp, int width)
352 const int width2 = width >> 1;
353 const int w2 = (width + 1) >> 1;
356 for (x = 0; x < width2; x++) {
358 temp[2 * x + 1] = b[x + w2];
363 b[0] = temp[0] - ((temp[1] + 1) >> 1);
364 for (x = 2; x < width - 1; x += 2) {
365 b[x] = temp[x] - ((temp[x - 1] + temp[x + 1] + 2) >> 2);
366 b[x - 1] = temp[x - 1] + ((b[x - 2] + b[x] + 1) >> 1);
369 b[x] = temp[x] - ((temp[x - 1] + 1) >> 1);
370 b[x - 1] = temp[x - 1] + ((b[x - 2] + b[x] + 1) >> 1);
372 b[x - 1] = temp[x - 1] + b[x - 2];
375 static void vertical_compose53iH0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
380 for (i = 0; i < width; i++)
381 b1[i] += (b0[i] + b2[i]) >> 1;
384 static void vertical_compose53iL0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
389 for (i = 0; i < width; i++)
390 b1[i] -= (b0[i] + b2[i] + 2) >> 2;
393 static void spatial_compose53i_buffered_init(DWTCompose *cs, slice_buffer *sb,
394 int height, int stride_line)
396 cs->b0 = slice_buffer_get_line(sb,
397 mirror(-1 - 1, height - 1) * stride_line);
398 cs->b1 = slice_buffer_get_line(sb, mirror(-1, height - 1) * stride_line);
402 static void spatial_compose53i_init(DWTCompose *cs, IDWTELEM *buffer,
403 int height, int stride)
405 cs->b0 = buffer + mirror(-1 - 1, height - 1) * stride;
406 cs->b1 = buffer + mirror(-1, height - 1) * stride;
410 static void spatial_compose53i_dy_buffered(DWTCompose *cs, slice_buffer *sb,
412 int width, int height,
417 IDWTELEM *b0 = cs->b0;
418 IDWTELEM *b1 = cs->b1;
419 IDWTELEM *b2 = slice_buffer_get_line(sb,
420 mirror(y + 1, height - 1) *
422 IDWTELEM *b3 = slice_buffer_get_line(sb,
423 mirror(y + 2, height - 1) *
426 if (y + 1 < (unsigned)height && y < (unsigned)height) {
429 for (x = 0; x < width; x++) {
430 b2[x] -= (b1[x] + b3[x] + 2) >> 2;
431 b1[x] += (b0[x] + b2[x]) >> 1;
434 if (y + 1 < (unsigned)height)
435 vertical_compose53iL0(b1, b2, b3, width);
436 if (y + 0 < (unsigned)height)
437 vertical_compose53iH0(b0, b1, b2, width);
440 if (y - 1 < (unsigned)height)
441 horizontal_compose53i(b0, temp, width);
442 if (y + 0 < (unsigned)height)
443 horizontal_compose53i(b1, temp, width);
450 static void spatial_compose53i_dy(DWTCompose *cs, IDWTELEM *buffer,
451 IDWTELEM *temp, int width, int height,
455 IDWTELEM *b0 = cs->b0;
456 IDWTELEM *b1 = cs->b1;
457 IDWTELEM *b2 = buffer + mirror(y + 1, height - 1) * stride;
458 IDWTELEM *b3 = buffer + mirror(y + 2, height - 1) * stride;
460 if (y + 1 < (unsigned)height)
461 vertical_compose53iL0(b1, b2, b3, width);
462 if (y + 0 < (unsigned)height)
463 vertical_compose53iH0(b0, b1, b2, width);
465 if (y - 1 < (unsigned)height)
466 horizontal_compose53i(b0, temp, width);
467 if (y + 0 < (unsigned)height)
468 horizontal_compose53i(b1, temp, width);
475 void ff_snow_horizontal_compose97i(IDWTELEM *b, IDWTELEM *temp, int width)
477 const int w2 = (width + 1) >> 1;
480 temp[0] = b[0] - ((3 * b[w2] + 2) >> 2);
481 for (x = 1; x < (width >> 1); x++) {
482 temp[2 * x] = b[x] - ((3 * (b[x + w2 - 1] + b[x + w2]) + 4) >> 3);
483 temp[2 * x - 1] = b[x + w2 - 1] - temp[2 * x - 2] - temp[2 * x];
486 temp[2 * x] = b[x] - ((3 * b[x + w2 - 1] + 2) >> 2);
487 temp[2 * x - 1] = b[x + w2 - 1] - temp[2 * x - 2] - temp[2 * x];
489 temp[2 * x - 1] = b[x + w2 - 1] - 2 * temp[2 * x - 2];
491 b[0] = temp[0] + ((2 * temp[0] + temp[1] + 4) >> 3);
492 for (x = 2; x < width - 1; x += 2) {
493 b[x] = temp[x] + ((4 * temp[x] + temp[x - 1] + temp[x + 1] + 8) >> 4);
494 b[x - 1] = temp[x - 1] + ((3 * (b[x - 2] + b[x])) >> 1);
497 b[x] = temp[x] + ((2 * temp[x] + temp[x - 1] + 4) >> 3);
498 b[x - 1] = temp[x - 1] + ((3 * (b[x - 2] + b[x])) >> 1);
500 b[x - 1] = temp[x - 1] + 3 * b[x - 2];
503 static void vertical_compose97iH0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
508 for (i = 0; i < width; i++)
509 b1[i] += (W_AM * (b0[i] + b2[i]) + W_AO) >> W_AS;
512 static void vertical_compose97iH1(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
517 for (i = 0; i < width; i++)
518 b1[i] -= (W_CM * (b0[i] + b2[i]) + W_CO) >> W_CS;
521 static void vertical_compose97iL0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
526 for (i = 0; i < width; i++)
527 b1[i] += (W_BM * (b0[i] + b2[i]) + 4 * b1[i] + W_BO) >> W_BS;
530 static void vertical_compose97iL1(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
535 for (i = 0; i < width; i++)
536 b1[i] -= (W_DM * (b0[i] + b2[i]) + W_DO) >> W_DS;
539 void ff_snow_vertical_compose97i(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
540 IDWTELEM *b3, IDWTELEM *b4, IDWTELEM *b5,
545 for (i = 0; i < width; i++) {
546 b4[i] -= (W_DM * (b3[i] + b5[i]) + W_DO) >> W_DS;
547 b3[i] -= (W_CM * (b2[i] + b4[i]) + W_CO) >> W_CS;
548 b2[i] += (W_BM * (b1[i] + b3[i]) + 4 * b2[i] + W_BO) >> W_BS;
549 b1[i] += (W_AM * (b0[i] + b2[i]) + W_AO) >> W_AS;
553 static void spatial_compose97i_buffered_init(DWTCompose *cs, slice_buffer *sb,
554 int height, int stride_line)
556 cs->b0 = slice_buffer_get_line(sb, mirror(-3 - 1, height - 1) * stride_line);
557 cs->b1 = slice_buffer_get_line(sb, mirror(-3, height - 1) * stride_line);
558 cs->b2 = slice_buffer_get_line(sb, mirror(-3 + 1, height - 1) * stride_line);
559 cs->b3 = slice_buffer_get_line(sb, mirror(-3 + 2, height - 1) * stride_line);
563 static void spatial_compose97i_init(DWTCompose *cs, IDWTELEM *buffer, int height,
566 cs->b0 = buffer + mirror(-3 - 1, height - 1) * stride;
567 cs->b1 = buffer + mirror(-3, height - 1) * stride;
568 cs->b2 = buffer + mirror(-3 + 1, height - 1) * stride;
569 cs->b3 = buffer + mirror(-3 + 2, height - 1) * stride;
573 static void spatial_compose97i_dy_buffered(SnowDWTContext *dsp, DWTCompose *cs,
574 slice_buffer * sb, IDWTELEM *temp,
575 int width, int height,
580 IDWTELEM *b0 = cs->b0;
581 IDWTELEM *b1 = cs->b1;
582 IDWTELEM *b2 = cs->b2;
583 IDWTELEM *b3 = cs->b3;
584 IDWTELEM *b4 = slice_buffer_get_line(sb,
585 mirror(y + 3, height - 1) *
587 IDWTELEM *b5 = slice_buffer_get_line(sb,
588 mirror(y + 4, height - 1) *
591 if (y > 0 && y + 4 < height) {
592 dsp->vertical_compose97i(b0, b1, b2, b3, b4, b5, width);
594 if (y + 3 < (unsigned)height)
595 vertical_compose97iL1(b3, b4, b5, width);
596 if (y + 2 < (unsigned)height)
597 vertical_compose97iH1(b2, b3, b4, width);
598 if (y + 1 < (unsigned)height)
599 vertical_compose97iL0(b1, b2, b3, width);
600 if (y + 0 < (unsigned)height)
601 vertical_compose97iH0(b0, b1, b2, width);
604 if (y - 1 < (unsigned)height)
605 dsp->horizontal_compose97i(b0, temp, width);
606 if (y + 0 < (unsigned)height)
607 dsp->horizontal_compose97i(b1, temp, width);
616 static void spatial_compose97i_dy(DWTCompose *cs, IDWTELEM *buffer,
617 IDWTELEM *temp, int width, int height,
621 IDWTELEM *b0 = cs->b0;
622 IDWTELEM *b1 = cs->b1;
623 IDWTELEM *b2 = cs->b2;
624 IDWTELEM *b3 = cs->b3;
625 IDWTELEM *b4 = buffer + mirror(y + 3, height - 1) * stride;
626 IDWTELEM *b5 = buffer + mirror(y + 4, height - 1) * stride;
628 if (y + 3 < (unsigned)height)
629 vertical_compose97iL1(b3, b4, b5, width);
630 if (y + 2 < (unsigned)height)
631 vertical_compose97iH1(b2, b3, b4, width);
632 if (y + 1 < (unsigned)height)
633 vertical_compose97iL0(b1, b2, b3, width);
634 if (y + 0 < (unsigned)height)
635 vertical_compose97iH0(b0, b1, b2, width);
637 if (y - 1 < (unsigned)height)
638 ff_snow_horizontal_compose97i(b0, temp, width);
639 if (y + 0 < (unsigned)height)
640 ff_snow_horizontal_compose97i(b1, temp, width);
649 void ff_spatial_idwt_buffered_init(DWTCompose *cs, slice_buffer *sb, int width,
650 int height, int stride_line, int type,
651 int decomposition_count)
654 for (level = decomposition_count - 1; level >= 0; level--) {
657 spatial_compose97i_buffered_init(cs + level, sb, height >> level,
658 stride_line << level);
661 spatial_compose53i_buffered_init(cs + level, sb, height >> level,
662 stride_line << level);
668 void ff_spatial_idwt_buffered_slice(SnowDWTContext *dsp, DWTCompose *cs,
669 slice_buffer *slice_buf, IDWTELEM *temp,
670 int width, int height, int stride_line,
671 int type, int decomposition_count, int y)
673 const int support = type == 1 ? 3 : 5;
678 for (level = decomposition_count - 1; level >= 0; level--)
679 while (cs[level].y <= FFMIN((y >> level) + support, height >> level)) {
682 spatial_compose97i_dy_buffered(dsp, cs + level, slice_buf, temp,
685 stride_line << level);
688 spatial_compose53i_dy_buffered(cs + level, slice_buf, temp,
691 stride_line << level);
697 static void ff_spatial_idwt_init(DWTCompose *cs, IDWTELEM *buffer, int width,
698 int height, int stride, int type,
699 int decomposition_count)
702 for (level = decomposition_count - 1; level >= 0; level--) {
705 spatial_compose97i_init(cs + level, buffer, height >> level,
709 spatial_compose53i_init(cs + level, buffer, height >> level,
716 static void ff_spatial_idwt_slice(DWTCompose *cs, IDWTELEM *buffer,
717 IDWTELEM *temp, int width, int height,
718 int stride, int type,
719 int decomposition_count, int y)
721 const int support = type == 1 ? 3 : 5;
726 for (level = decomposition_count - 1; level >= 0; level--)
727 while (cs[level].y <= FFMIN((y >> level) + support, height >> level)) {
730 spatial_compose97i_dy(cs + level, buffer, temp, width >> level,
731 height >> level, stride << level);
734 spatial_compose53i_dy(cs + level, buffer, temp, width >> level,
735 height >> level, stride << level);
741 void ff_spatial_idwt(IDWTELEM *buffer, IDWTELEM *temp, int width, int height,
742 int stride, int type, int decomposition_count)
744 DWTCompose cs[MAX_DECOMPOSITIONS];
746 ff_spatial_idwt_init(cs, buffer, width, height, stride, type,
747 decomposition_count);
748 for (y = 0; y < height; y += 4)
749 ff_spatial_idwt_slice(cs, buffer, temp, width, height, stride, type,
750 decomposition_count, y);
753 static inline int w_c(struct MpegEncContext *v, uint8_t *pix1, uint8_t *pix2, ptrdiff_t line_size,
754 int w, int h, int type)
757 const int dec_count = w == 8 ? 3 : 4;
758 int tmp[32 * 32], tmp2[32];
760 static const int scale[2][2][4][4] = {
763 { 268, 239, 239, 213 },
764 { 0, 224, 224, 152 },
765 { 0, 135, 135, 110 },
767 { // 9/7 16x16 or 32x32 dec=4
768 { 344, 310, 310, 280 },
769 { 0, 320, 320, 228 },
770 { 0, 175, 175, 136 },
771 { 0, 129, 129, 102 },
776 { 275, 245, 245, 218 },
777 { 0, 230, 230, 156 },
778 { 0, 138, 138, 113 },
780 { // 5/3 16x16 or 32x32 dec=4
781 { 352, 317, 317, 286 },
782 { 0, 328, 328, 233 },
783 { 0, 180, 180, 140 },
784 { 0, 132, 132, 105 },
789 for (i = 0; i < h; i++) {
790 for (j = 0; j < w; j += 4) {
791 tmp[32 * i + j + 0] = (pix1[j + 0] - pix2[j + 0]) << 4;
792 tmp[32 * i + j + 1] = (pix1[j + 1] - pix2[j + 1]) << 4;
793 tmp[32 * i + j + 2] = (pix1[j + 2] - pix2[j + 2]) << 4;
794 tmp[32 * i + j + 3] = (pix1[j + 3] - pix2[j + 3]) << 4;
800 ff_spatial_dwt(tmp, tmp2, w, h, 32, type, dec_count);
804 for (level = 0; level < dec_count; level++)
805 for (ori = level ? 1 : 0; ori < 4; ori++) {
806 int size = w >> (dec_count - level);
807 int sx = (ori & 1) ? size : 0;
808 int stride = 32 << (dec_count - level);
809 int sy = (ori & 2) ? stride >> 1 : 0;
811 for (i = 0; i < size; i++)
812 for (j = 0; j < size; j++) {
813 int v = tmp[sx + sy + i * stride + j] *
814 scale[type][dec_count - 3][level][ori];
822 static int w53_8_c(struct MpegEncContext *v, uint8_t *pix1, uint8_t *pix2, ptrdiff_t line_size, int h)
824 return w_c(v, pix1, pix2, line_size, 8, h, 1);
827 static int w97_8_c(struct MpegEncContext *v, uint8_t *pix1, uint8_t *pix2, ptrdiff_t line_size, int h)
829 return w_c(v, pix1, pix2, line_size, 8, h, 0);
832 static int w53_16_c(struct MpegEncContext *v, uint8_t *pix1, uint8_t *pix2, ptrdiff_t line_size, int h)
834 return w_c(v, pix1, pix2, line_size, 16, h, 1);
837 static int w97_16_c(struct MpegEncContext *v, uint8_t *pix1, uint8_t *pix2, ptrdiff_t line_size, int h)
839 return w_c(v, pix1, pix2, line_size, 16, h, 0);
842 int ff_w53_32_c(struct MpegEncContext *v, uint8_t *pix1, uint8_t *pix2, ptrdiff_t line_size, int h)
844 return w_c(v, pix1, pix2, line_size, 32, h, 1);
847 int ff_w97_32_c(struct MpegEncContext *v, uint8_t *pix1, uint8_t *pix2, ptrdiff_t line_size, int h)
849 return w_c(v, pix1, pix2, line_size, 32, h, 0);
852 void ff_dsputil_init_dwt(MECmpContext *c)
854 c->w53[0] = w53_16_c;
856 c->w97[0] = w97_16_c;
860 void ff_dwt_init(SnowDWTContext *c)
862 c->vertical_compose97i = ff_snow_vertical_compose97i;
863 c->horizontal_compose97i = ff_snow_horizontal_compose97i;
864 c->inner_add_yblock = ff_snow_inner_add_yblock;