1 |
|
/***************************************************************************** |
2 |
|
* |
3 |
|
* XVID MPEG-4 VIDEO CODEC |
4 |
|
* - Postprocessing functions - |
5 |
|
* |
6 |
|
* Copyright(C) 2003 Michael Militzer <isibaar@xvid.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 |
|
* $Id$ |
23 |
|
* |
24 |
|
****************************************************************************/ |
25 |
|
|
26 |
|
#include <stdlib.h> |
27 |
|
#include <string.h> |
28 |
|
#include <math.h> |
29 |
|
|
30 |
|
#include "../portab.h" |
31 |
|
#include "../global.h" |
32 |
|
#include "image.h" |
33 |
|
#include "../utils/emms.h" |
34 |
|
#include "postprocessing.h" |
35 |
|
|
36 |
|
/* Filtering thresholds */ |
37 |
|
|
38 |
|
#define THR1 2 |
39 |
|
#define THR2 6 |
40 |
|
|
41 |
|
/* Some useful (and fast) macros |
42 |
|
Note that the MIN/MAX macros assume signed shift - if your compiler |
43 |
|
doesn't do signed shifts, use the default MIN/MAX macros from global.h */ |
44 |
|
|
45 |
|
#define FAST_MAX(x,y) ((x) - ((((x) - (y))>>(32 - 1)) & ((x) - (y)))) |
46 |
|
#define FAST_MIN(x,y) ((x) + ((((y) - (x))>>(32 - 1)) & ((y) - (x)))) |
47 |
|
#define FAST_ABS(x) ((((int)(x)) >> 31) ^ ((int)(x))) - (((int)(x)) >> 31) |
48 |
|
#define ABS(X) (((X)>0)?(X):-(X)) |
49 |
|
|
50 |
|
void init_postproc(void) |
51 |
|
{ |
52 |
|
init_deblock(); |
53 |
|
init_noise(); |
54 |
|
} |
55 |
|
|
56 |
|
void |
57 |
|
image_postproc(IMAGE * img, int edged_width, |
58 |
|
const MACROBLOCK * mbs, int mb_width, int mb_height, int mb_stride, |
59 |
|
int flags, int frame_num) |
60 |
|
{ |
61 |
|
const int edged_width2 = edged_width /2; |
62 |
|
int i,j; |
63 |
|
int quant; |
64 |
|
|
65 |
|
/* luma: j,i in block units */ |
66 |
|
if ((flags & XVID_DEBLOCKY)) |
67 |
|
{ |
68 |
|
for (j = 1; j < mb_height*2; j++) /* horizontal deblocking */ |
69 |
|
for (i = 0; i < mb_width*2; i++) |
70 |
|
{ |
71 |
|
quant = mbs[(j+0)/2*mb_stride + (i/2)].quant; |
72 |
|
deblock8x8_h(img->y + j*8*edged_width + i*8, edged_width, quant); |
73 |
|
} |
74 |
|
|
75 |
|
for (j = 0; j < mb_height*2; j++) /* vertical deblocking */ |
76 |
|
for (i = 1; i < mb_width*2; i++) |
77 |
|
{ |
78 |
|
quant = mbs[(j+0)/2*mb_stride + (i/2)].quant; |
79 |
|
deblock8x8_v(img->y + j*8*edged_width + i*8, edged_width, quant); |
80 |
|
} |
81 |
|
} |
82 |
|
|
83 |
|
|
84 |
|
/* chroma */ |
85 |
|
if ((flags & XVID_DEBLOCKUV)) |
86 |
|
{ |
87 |
|
for (j = 1; j < mb_height; j++) /* horizontal deblocking */ |
88 |
|
for (i = 0; i < mb_width; i++) |
89 |
|
{ |
90 |
|
quant = mbs[(j+0)*mb_stride + i].quant; |
91 |
|
deblock8x8_h(img->u + j*8*edged_width2 + i*8, edged_width2, quant); |
92 |
|
deblock8x8_h(img->v + j*8*edged_width2 + i*8, edged_width2, quant); |
93 |
|
} |
94 |
|
|
95 |
|
for (j = 0; j < mb_height; j++) /* vertical deblocking */ |
96 |
|
for (i = 1; i < mb_width; i++) |
97 |
|
{ |
98 |
|
quant = mbs[(j+0)*mb_stride + i].quant; |
99 |
|
deblock8x8_v(img->u + j*8*edged_width2 + i*8, edged_width2, quant); |
100 |
|
deblock8x8_v(img->v + j*8*edged_width2 + i*8, edged_width2, quant); |
101 |
|
} |
102 |
|
} |
103 |
|
|
104 |
|
if ((flags & XVID_FILMEFFECT)) |
105 |
|
{ |
106 |
|
add_noise(img->y, img->y, edged_width, mb_width*16, mb_height*16, frame_num % 3); |
107 |
|
} |
108 |
|
} |
109 |
|
|
110 |
|
/******************************************************************************/ |
111 |
|
|
112 |
|
static int8_t xvid_thresh_tbl[510]; |
113 |
|
static int8_t xvid_abs_tbl[510]; |
114 |
|
|
115 |
|
void init_deblock(void) |
116 |
|
{ |
117 |
|
int i; |
118 |
|
|
119 |
|
for(i = -255; i < 256; i++) { |
120 |
|
xvid_thresh_tbl[i + 255] = 0; |
121 |
|
if(ABS(i) < THR1) |
122 |
|
xvid_thresh_tbl[i + 255] = 1; |
123 |
|
xvid_abs_tbl[i + 255] = ABS(i); |
124 |
|
} |
125 |
|
} |
126 |
|
|
127 |
|
#define LOAD_DATA_HOR(x) \ |
128 |
|
/* Load pixel addresses and data for filtering */ \ |
129 |
|
s[0] = *(v[0] = img - 5*stride + x); \ |
130 |
|
s[1] = *(v[1] = img - 4*stride + x); \ |
131 |
|
s[2] = *(v[2] = img - 3*stride + x); \ |
132 |
|
s[3] = *(v[3] = img - 2*stride + x); \ |
133 |
|
s[4] = *(v[4] = img - 1*stride + x); \ |
134 |
|
s[5] = *(v[5] = img + 0*stride + x); \ |
135 |
|
s[6] = *(v[6] = img + 1*stride + x); \ |
136 |
|
s[7] = *(v[7] = img + 2*stride + x); \ |
137 |
|
s[8] = *(v[8] = img + 3*stride + x); \ |
138 |
|
s[9] = *(v[9] = img + 4*stride + x); |
139 |
|
|
140 |
|
#define LOAD_DATA_VER(x) \ |
141 |
|
/* Load pixel addresses and data for filtering */ \ |
142 |
|
s[0] = *(v[0] = img + x*stride - 5); \ |
143 |
|
s[1] = *(v[1] = img + x*stride - 4); \ |
144 |
|
s[2] = *(v[2] = img + x*stride - 3); \ |
145 |
|
s[3] = *(v[3] = img + x*stride - 2); \ |
146 |
|
s[4] = *(v[4] = img + x*stride - 1); \ |
147 |
|
s[5] = *(v[5] = img + x*stride + 0); \ |
148 |
|
s[6] = *(v[6] = img + x*stride + 1); \ |
149 |
|
s[7] = *(v[7] = img + x*stride + 2); \ |
150 |
|
s[8] = *(v[8] = img + x*stride + 3); \ |
151 |
|
s[9] = *(v[9] = img + x*stride + 4); |
152 |
|
|
153 |
|
#define APPLY_FILTER_CORE \ |
154 |
|
/* First, decide whether to use default or DC-offset mode */ \ |
155 |
|
\ |
156 |
|
eq_cnt = 0; \ |
157 |
|
\ |
158 |
|
eq_cnt += xvid_thresh_tbl[s[0] - s[1] + 255]; \ |
159 |
|
eq_cnt += xvid_thresh_tbl[s[1] - s[2] + 255]; \ |
160 |
|
eq_cnt += xvid_thresh_tbl[s[2] - s[3] + 255]; \ |
161 |
|
eq_cnt += xvid_thresh_tbl[s[3] - s[4] + 255]; \ |
162 |
|
eq_cnt += xvid_thresh_tbl[s[4] - s[5] + 255]; \ |
163 |
|
eq_cnt += xvid_thresh_tbl[s[5] - s[6] + 255]; \ |
164 |
|
eq_cnt += xvid_thresh_tbl[s[6] - s[7] + 255]; \ |
165 |
|
eq_cnt += xvid_thresh_tbl[s[7] - s[8] + 255]; \ |
166 |
|
\ |
167 |
|
if(eq_cnt < THR2) { /* Default mode */ \ |
168 |
|
int a30, a31, a32; \ |
169 |
|
int diff, limit; \ |
170 |
|
\ |
171 |
|
if(xvid_abs_tbl[(s[4] - s[5]) + 255] < quant) { \ |
172 |
|
a30 = ((s[3]<<1) - s[4] * 5 + s[5] * 5 - (s[6]<<1)); \ |
173 |
|
a31 = ((s[1]<<1) - s[2] * 5 + s[3] * 5 - (s[4]<<1)); \ |
174 |
|
a32 = ((s[5]<<1) - s[6] * 5 + s[7] * 5 - (s[8]<<1)); \ |
175 |
|
\ |
176 |
|
diff = (5 * ((SIGN(a30) * MIN(xvid_abs_tbl[a30 + 255], MIN(xvid_abs_tbl[a31 + 255], xvid_abs_tbl[a32 + 255]))) - a30) + 32) >> 6; \ |
177 |
|
limit = (s[4] - s[5]) / 2; \ |
178 |
|
\ |
179 |
|
if (limit > 0) \ |
180 |
|
diff = (diff < 0) ? 0 : ((diff > limit) ? limit : diff); \ |
181 |
|
else \ |
182 |
|
diff = (diff > 0) ? 0 : ((diff < limit) ? limit : diff); \ |
183 |
|
\ |
184 |
|
*v[4] -= diff; \ |
185 |
|
*v[5] += diff; \ |
186 |
|
} \ |
187 |
|
} \ |
188 |
|
else { /* DC-offset mode */ \ |
189 |
|
uint8_t p0, p9; \ |
190 |
|
int min, max; \ |
191 |
|
\ |
192 |
|
/* Now decide whether to apply smoothing filter or not */ \ |
193 |
|
max = FAST_MAX(s[1], FAST_MAX(s[2], FAST_MAX(s[3], FAST_MAX(s[4], FAST_MAX(s[5], FAST_MAX(s[6], FAST_MAX(s[7], s[8]))))))); \ |
194 |
|
min = FAST_MIN(s[1], FAST_MIN(s[2], FAST_MIN(s[3], FAST_MIN(s[4], FAST_MIN(s[5], FAST_MIN(s[6], FAST_MIN(s[7], s[8]))))))); \ |
195 |
|
\ |
196 |
|
if(((max-min)) < 2*quant) { \ |
197 |
|
\ |
198 |
|
/* Choose edge pixels */ \ |
199 |
|
p0 = (xvid_abs_tbl[(s[1] - s[0]) + 255] < quant) ? s[0] : s[1]; \ |
200 |
|
p9 = (xvid_abs_tbl[(s[8] - s[9]) + 255] < quant) ? s[9] : s[8]; \ |
201 |
|
\ |
202 |
|
*v[1] = (uint8_t) ((6*p0 + (s[1]<<2) + (s[2]<<1) + (s[3]<<1) + s[4] + s[5] + 8) >> 4); \ |
203 |
|
*v[2] = (uint8_t) (((p0<<2) + (s[1]<<1) + (s[2]<<2) + (s[3]<<1) + (s[4]<<1) + s[5] + s[6] + 8) >> 4); \ |
204 |
|
*v[3] = (uint8_t) (((p0<<1) + (s[1]<<1) + (s[2]<<1) + (s[3]<<2) + (s[4]<<1) + (s[5]<<1) + s[6] + s[7] + 8) >> 4); \ |
205 |
|
*v[4] = (uint8_t) ((p0 + s[1] + (s[2]<<1) + (s[3]<<1) + (s[4]<<2) + (s[5]<<1) + (s[6]<<1) + s[7] + s[8] + 8) >> 4); \ |
206 |
|
*v[5] = (uint8_t) ((s[1] + s[2] + (s[3]<<1) + (s[4]<<1) + (s[5]<<2) + (s[6]<<1) + (s[7]<<1) + s[8] + p9 + 8) >> 4); \ |
207 |
|
*v[6] = (uint8_t) ((s[2] + s[3] + (s[4]<<1) + (s[5]<<1) + (s[6]<<2) + (s[7]<<1) + (s[8]<<1) + (p9<<1) + 8) >> 4); \ |
208 |
|
*v[7] = (uint8_t) ((s[3] + s[4] + (s[5]<<1) + (s[6]<<1) + (s[7]<<2) + (s[8]<<1) + (p9<<2) + 8) >> 4); \ |
209 |
|
*v[8] = (uint8_t) ((s[4] + s[5] + (s[6]<<1) + (s[7]<<1) + (s[8]<<2) + 6*p9 + 8) >> 4); \ |
210 |
|
} \ |
211 |
|
} |
212 |
|
|
213 |
|
void deblock8x8_h(uint8_t *img, int stride, int quant) |
214 |
|
{ |
215 |
|
int eq_cnt; |
216 |
|
uint8_t *v[10]; |
217 |
|
int32_t s[10]; |
218 |
|
|
219 |
|
LOAD_DATA_HOR(0) |
220 |
|
APPLY_FILTER_CORE |
221 |
|
|
222 |
|
LOAD_DATA_HOR(1) |
223 |
|
APPLY_FILTER_CORE |
224 |
|
|
225 |
|
LOAD_DATA_HOR(2) |
226 |
|
APPLY_FILTER_CORE |
227 |
|
|
228 |
|
LOAD_DATA_HOR(3) |
229 |
|
APPLY_FILTER_CORE |
230 |
|
|
231 |
|
LOAD_DATA_HOR(4) |
232 |
|
APPLY_FILTER_CORE |
233 |
|
|
234 |
|
LOAD_DATA_HOR(5) |
235 |
|
APPLY_FILTER_CORE |
236 |
|
|
237 |
|
LOAD_DATA_HOR(6) |
238 |
|
APPLY_FILTER_CORE |
239 |
|
|
240 |
|
LOAD_DATA_HOR(7) |
241 |
|
APPLY_FILTER_CORE |
242 |
|
} |
243 |
|
|
244 |
|
|
245 |
|
void deblock8x8_v(uint8_t *img, int stride, int quant) |
246 |
|
{ |
247 |
|
int eq_cnt; |
248 |
|
uint8_t *v[10]; |
249 |
|
int s[10]; |
250 |
|
|
251 |
|
LOAD_DATA_VER(0) |
252 |
|
APPLY_FILTER_CORE |
253 |
|
|
254 |
|
LOAD_DATA_VER(1) |
255 |
|
APPLY_FILTER_CORE |
256 |
|
|
257 |
|
LOAD_DATA_VER(2) |
258 |
|
APPLY_FILTER_CORE |
259 |
|
|
260 |
|
LOAD_DATA_VER(3) |
261 |
|
APPLY_FILTER_CORE |
262 |
|
|
263 |
|
LOAD_DATA_VER(4) |
264 |
|
APPLY_FILTER_CORE |
265 |
|
|
266 |
|
LOAD_DATA_VER(5) |
267 |
|
APPLY_FILTER_CORE |
268 |
|
|
269 |
|
LOAD_DATA_VER(6) |
270 |
|
APPLY_FILTER_CORE |
271 |
|
|
272 |
|
LOAD_DATA_VER(7) |
273 |
|
APPLY_FILTER_CORE |
274 |
|
} |
275 |
|
|
276 |
|
/****************************************************************************** |
277 |
|
* * |
278 |
|
* Noise code below taken from MPlayer: http://www.mplayerhq.hu/ * |
279 |
|
* Copyright (C) 2002 Michael Niedermayer <michaelni@gmx.at> * |
280 |
|
* * |
281 |
|
******************************************************************************/ |
282 |
|
|
283 |
|
#define MAX_NOISE 4096 |
284 |
|
#define MAX_SHIFT 1024 |
285 |
|
#define MAX_RES (MAX_NOISE - MAX_SHIFT) |
286 |
|
|
287 |
|
#define RAND_N(range) ((int) ((double)range * rand() / (RAND_MAX + 1.0))) |
288 |
|
|
289 |
|
#define STRENGTH 13 |
290 |
|
|
291 |
|
static int8_t xvid_noise[MAX_NOISE * sizeof(int8_t)]; |
292 |
|
static int8_t *xvid_prev_shift[MAX_RES][3]; |
293 |
|
|
294 |
|
void init_noise(void) |
295 |
|
{ |
296 |
|
int i, j; |
297 |
|
int patt[4] = { -1,0,1,0 }; |
298 |
|
|
299 |
|
emms(); |
300 |
|
|
301 |
|
srand(123457); |
302 |
|
|
303 |
|
for(i = 0, j = 0; i < MAX_NOISE; i++, j++) |
304 |
|
{ |
305 |
|
double x1, x2, w, y1; |
306 |
|
|
307 |
|
do { |
308 |
|
x1 = 2.0 * rand() / (float) RAND_MAX - 1.0; |
309 |
|
x2 = 2.0 * rand() / (float) RAND_MAX - 1.0; |
310 |
|
w = x1 * x1 + x2 * x2; |
311 |
|
} while (w >= 1.0); |
312 |
|
|
313 |
|
w = sqrt((-2.0 * log(w)) / w); |
314 |
|
y1 = x1 * w; |
315 |
|
y1 *= STRENGTH / sqrt(3.0); |
316 |
|
|
317 |
|
y1 /= 2; |
318 |
|
y1 += patt[j%4] * STRENGTH * 0.35; |
319 |
|
|
320 |
|
if (y1 < -128) { |
321 |
|
y1=-128; |
322 |
|
} |
323 |
|
else if (y1 > 127) { |
324 |
|
y1= 127; |
325 |
|
} |
326 |
|
|
327 |
|
y1 /= 3.0; |
328 |
|
xvid_noise[i] = (int) y1; |
329 |
|
|
330 |
|
if (RAND_N(6) == 0) { |
331 |
|
j--; |
332 |
|
} |
333 |
|
} |
334 |
|
|
335 |
|
for (i = 0; i < MAX_RES; i++) |
336 |
|
for (j = 0; j < 3; j++) { |
337 |
|
xvid_prev_shift[i][j] = xvid_noise + (rand() & (MAX_SHIFT - 1)); |
338 |
|
} |
339 |
|
} |
340 |
|
|
341 |
|
void add_noise(uint8_t *dst, uint8_t *src, int stride, int width, int height, int shiftptr) |
342 |
|
{ |
343 |
|
int x, y; |
344 |
|
int shift = 0; |
345 |
|
|
346 |
|
for(y = 0; y < height; y++) |
347 |
|
{ |
348 |
|
int8_t *src2 = (int8_t *) src; |
349 |
|
|
350 |
|
shift = rand() & (MAX_SHIFT - 1); |
351 |
|
|
352 |
|
shift &= ~7; |
353 |
|
for(x = 0; x < width; x++) |
354 |
|
{ |
355 |
|
const int n = xvid_prev_shift[y][0][x] + xvid_prev_shift[y][1][x] + |
356 |
|
xvid_prev_shift[y][2][x]; |
357 |
|
|
358 |
|
dst[x] = src2[x] + ((n * src2[x]) >> 7); |
359 |
|
} |
360 |
|
|
361 |
|
xvid_prev_shift[y][shiftptr] = xvid_noise + shift; |
362 |
|
|
363 |
|
dst += stride; |
364 |
|
src += stride; |
365 |
|
} |
366 |
|
} |