aboutsummaryrefslogtreecommitdiffhomepage
path: root/libvo/jpeg_enc.c
blob: eacd68e5bbe2ad8857b997f2daa9a2ce4bf635b0 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
/* Straightforward (to be) optimized JPEG encoder for the YUV422 format 
 * based on mjpeg code from ffmpeg. 
 *
 * Copyright (c) 2002, Rik Snel
 * Parts from ffmpeg Copyright (c) 2000-2002 Fabrice Bellard
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 *
 * For an excellent introduction to the JPEG format, see:
 * http://www.ece.purdue.edu/~bouman/grad-labs/lab8/pdf/lab.pdf
 */



#include <sys/types.h>
#include <stdio.h>
#include <stdlib.h>
#include "config.h"
#include "mp_msg.h"
/* We need this #define because we need ../libavcodec/common.h to #define 
 * be2me_32, otherwise the linker will complain that it doesn't exist */
#define HAVE_AV_CONFIG_H
#include "libavcodec/avcodec.h"
#include "libavcodec/dsputil.h"
#include "libavcodec/mpegvideo.h"
#include "libavcodec/mjpegenc.h"

#include "jpeg_enc.h"

extern int avcodec_initialized;


/* Begin excessive code duplication ************************************/
/* Code coming from mpegvideo.c and mjpeg.c in ../libavcodec ***********/

static const unsigned short aanscales[64] = {
    /* precomputed values scaled up by 14 bits */
    16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
    22725, 31521, 29692, 26722, 22725, 17855, 12299,  6270,
    21407, 29692, 27969, 25172, 21407, 16819, 11585,  5906,
    19266, 26722, 25172, 22654, 19266, 15137, 10426,  5315,
    16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
    12873, 17855, 16819, 15137, 12873, 10114,  6967,  3552,
    8867, 12299, 11585, 10426,  8867,  6967,  4799,  2446,
    4520,  6270,  5906,  5315,  4520,  3552,  2446,  1247
};

static void convert_matrix(MpegEncContext *s, int (*qmat)[64], 
		uint16_t (*qmat16)[2][64], const uint16_t *quant_matrix,
		int bias, int qmin, int qmax)
{
    int qscale;

    for(qscale=qmin; qscale<=qmax; qscale++){
        int i;
	if (s->dsp.fdct == ff_jpeg_fdct_islow) {
		for (i = 0; i < 64; i++) {
			const int j = s->dsp.idct_permutation[i];
			/* 16    <= qscale * quant_matrix[i] <= 7905 
			 * 19952 <= aanscales[i] *  \
			 * 	        qscale * quant_matrix[i]     <= 205026 
			 * (1<<36)/19952 >= (1<<36)/(aanscales[i] * \
			 * 	qscale * quant_matrix[i]) >= (1<<36)/249205025
			 * 3444240       >= (1<<36)/(aanscales[i] *
			 *      qscale * quant_matrix[i]) >= 275              */
			qmat[qscale][i] = (int)((UINT64_C(1) << (QMAT_SHIFT-3))/
					(qscale * quant_matrix[j]));
		}
	} else if (s->dsp.fdct == fdct_ifast) {
            for(i=0;i<64;i++) {
                const int j = s->dsp.idct_permutation[i];
                /* 16 <= qscale * quant_matrix[i] <= 7905 */
                /* 19952         <= aanscales[i] * qscale * quant_matrix[i]           <= 249205026 */
                /* (1<<36)/19952 >= (1<<36)/(aanscales[i] * qscale * quant_matrix[i]) >= (1<<36)/249205026 */
                /* 3444240       >= (1<<36)/(aanscales[i] * qscale * quant_matrix[i]) >= 275 */
                
                qmat[qscale][i] = (int)((UINT64_C(1) << (QMAT_SHIFT + 11)) / 
                                (aanscales[i] * qscale * quant_matrix[j]));
            }
        } else {
            for(i=0;i<64;i++) {
		const int j = s->dsp.idct_permutation[i];
                /* We can safely suppose that 16 <= quant_matrix[i] <= 255
                   So 16           <= qscale * quant_matrix[i]             <= 7905
                   so (1<<19) / 16 >= (1<<19) / (qscale * quant_matrix[i]) >= (1<<19) / 7905
                   so 32768        >= (1<<19) / (qscale * quant_matrix[i]) >= 67
                */
                qmat  [qscale][i] = (int)((UINT64_C(1) << QMAT_SHIFT_MMX) / (qscale * quant_matrix[j]));
                qmat16[qscale][0][i] = (1 << QMAT_SHIFT_MMX) / (qscale * quant_matrix[j]);

                if(qmat16[qscale][0][i]==0 || qmat16[qscale][0][i]==128*256) qmat16[qscale][0][i]=128*256-1;
                qmat16[qscale][1][i]= ROUNDED_DIV(bias<<(16-QUANT_BIAS_SHIFT), qmat16[qscale][0][i]);
            }
        }
    }
}

static inline void encode_dc(MpegEncContext *s, int val, 
                             uint8_t *huff_size, uint16_t *huff_code)
{
    int mant, nbits;

    if (val == 0) {
        put_bits(&s->pb, huff_size[0], huff_code[0]);
    } else {
        mant = val;
        if (val < 0) {
            val = -val;
            mant--;
        }
        
        /* compute the log (XXX: optimize) */
        nbits = 0;
        while (val != 0) {
            val = val >> 1;
            nbits++;
        }
            
        put_bits(&s->pb, huff_size[nbits], huff_code[nbits]);
        
        put_bits(&s->pb, nbits, mant & ((1 << nbits) - 1));
    }
}

static void encode_block(MpegEncContext *s, DCTELEM *block, int n)
{
    int mant, nbits, code, i, j;
    int component, dc, run, last_index, val;
    MJpegContext *m = s->mjpeg_ctx;
    uint8_t *huff_size_ac;
    uint16_t *huff_code_ac;
    
    /* DC coef */
    component = (n <= 3 ? 0 : n - 4 + 1);
    dc = block[0]; /* overflow is impossible */
    val = dc - s->last_dc[component];
    if (n < 4) {
        encode_dc(s, val, m->huff_size_dc_luminance, m->huff_code_dc_luminance);
        huff_size_ac = m->huff_size_ac_luminance;
        huff_code_ac = m->huff_code_ac_luminance;
    } else {
        encode_dc(s, val, m->huff_size_dc_chrominance, m->huff_code_dc_chrominance);
        huff_size_ac = m->huff_size_ac_chrominance;
        huff_code_ac = m->huff_code_ac_chrominance;
    }
    s->last_dc[component] = dc;
    
    /* AC coefs */
    
    run = 0;
    last_index = s->block_last_index[n];
    for(i=1;i<=last_index;i++) {
        j = s->intra_scantable.permutated[i];
        val = block[j];
        if (val == 0) {
            run++;
        } else {
            while (run >= 16) {
                put_bits(&s->pb, huff_size_ac[0xf0], huff_code_ac[0xf0]);
                run -= 16;
            }
            mant = val;
            if (val < 0) {
                val = -val;
                mant--;
            }
            
            /* compute the log (XXX: optimize) */
            nbits = 0;
            while (val != 0) {
                val = val >> 1;
                nbits++;
            }
            code = (run << 4) | nbits;

            put_bits(&s->pb, huff_size_ac[code], huff_code_ac[code]);
        
            put_bits(&s->pb, nbits, mant & ((1 << nbits) - 1));
            run = 0;
        }
    }

    /* output EOB only if not already 64 values */
    if (last_index < 63 || run != 0)
        put_bits(&s->pb, huff_size_ac[0], huff_code_ac[0]);
}

static inline void clip_coeffs(MpegEncContext *s, DCTELEM *block, int last_index)
{
    int i;
    const int maxlevel= s->max_qcoeff;
    const int minlevel= s->min_qcoeff;

    for(i=0; i<=last_index; i++){
        const int j = s->intra_scantable.permutated[i];
        int level = block[j];
       
        if     (level>maxlevel) level=maxlevel;
        else if(level<minlevel) level=minlevel;
        block[j]= level;
    }
}

/* End excessive code duplication **************************************/

/* this function is a reproduction of the one in mjpeg, it includes two
 * changes, it allows for black&white encoding (it skips the U and V
 * macroblocks and it outputs the huffman code for 'no change' (dc) and
 * 'all zero' (ac)) and it takes 4 macroblocks (422) instead of 6 (420) */
static void zr_mjpeg_encode_mb(jpeg_enc_t *j) {

	MJpegContext *m = j->s->mjpeg_ctx;

	encode_block(j->s, j->s->block[0], 0);
	encode_block(j->s, j->s->block[1], 1);
	if (j->bw) {
		/* U */
		put_bits(&j->s->pb, m->huff_size_dc_chrominance[0],
				m->huff_code_dc_chrominance[0]);
		put_bits(&j->s->pb, m->huff_size_ac_chrominance[0],
				m->huff_code_ac_chrominance[0]);
		/* V */
		put_bits(&j->s->pb, m->huff_size_dc_chrominance[0],
				m->huff_code_dc_chrominance[0]);
		put_bits(&j->s->pb, m->huff_size_ac_chrominance[0],
				m->huff_code_ac_chrominance[0]);
    	} else {
		/* we trick encode_block here so that it uses
		 * chrominance huffman tables instead of luminance ones 
		 * (see the effect of second argument of encode_block) */
		encode_block(j->s, j->s->block[2], 4); 
		encode_block(j->s, j->s->block[3], 5);
    	}
}

/* this function can take all kinds of YUV colorspaces
 * YV12, YVYU, UYVY. The necesary parameters must be set up by the caller
 * y_ps means "y pixel size", y_rs means "y row size".
 * For YUYV, for example, is u_buf = y_buf + 1, v_buf = y_buf + 3, 
 * y_ps = 2, u_ps = 4, v_ps = 4, y_rs = u_rs = v_rs.
 *
 *  The actual buffers must be passed with mjpeg_encode_frame, this is
 *  to make it possible to call encode on the buffer provided by the
 *  codec in draw_frame.
 *  
 * The data is straightened out at the moment it is put in DCT
 * blocks, there are therefore no spurious memcopies involved */
/* Notice that w must be a multiple of 16 and h must be a multiple of 8 */
/* We produce YUV422 jpegs, the colors must be subsampled horizontally,
 * if the colors are also subsampled vertically, then this function
 * performs cheap upsampling (better solution will be: a DCT that is
 * optimized in the case that every two rows are the same) */
/* cu = 0 means 'No cheap upsampling'
 * cu = 1 means 'perform cheap upsampling' */
/* The encoder doesn't know anything about interlacing, the halve height
 * needs to be passed and the double rowstride. Which field gets encoded
 * is decided by what buffers are passed to mjpeg_encode_frame */
jpeg_enc_t *jpeg_enc_init(int w, int h, int y_psize, int y_rsize, 
		int u_psize, int u_rsize, int v_psize, int v_rsize,
		int cu, int q, int b) {
	jpeg_enc_t *j;
	int i = 0;
	mp_msg(MSGT_VO, MSGL_V, "JPEnc init: %dx%d %d %d %d %d %d %d\n",
			w, h, y_psize, y_rsize, u_psize, 
			u_rsize, v_psize, v_rsize);

	j = av_malloc(sizeof(jpeg_enc_t));
	if (j == NULL) return NULL;

	j->s = av_malloc(sizeof(MpegEncContext));
	memset(j->s,0x00,sizeof(MpegEncContext));
	if (j->s == NULL) {
		av_free(j);
		return NULL;
	}

	/* info on how to access the pixels */
	j->y_ps = y_psize; 
	j->u_ps = u_psize; 
	j->v_ps = v_psize;
	j->y_rs = y_rsize; 
	j->u_rs = u_rsize; 
	j->v_rs = v_rsize;

	j->s->width = w;
	j->s->height = h;
	j->s->qscale = q;

	j->s->out_format = FMT_MJPEG;
	j->s->intra_only = 1;
	j->s->encoding = 1;
	j->s->pict_type = FF_I_TYPE;
	j->s->y_dc_scale = 8;
	j->s->c_dc_scale = 8;

	//FIXME j->s->mjpeg_write_tables = 1;
	j->s->mjpeg_vsample[0] = 1;
	j->s->mjpeg_vsample[1] = 1;
	j->s->mjpeg_vsample[2] = 1;
	j->s->mjpeg_hsample[0] = 2;
	j->s->mjpeg_hsample[1] = 1;
	j->s->mjpeg_hsample[2] = 1;

	j->cheap_upsample = cu;
	j->bw = b;

	/* if libavcodec is used by the decoder then we must not
	 * initialize again, but if it is not initialized then we must
	 * initialize it here. */
	if (!avcodec_initialized) {
		/* we need to initialize libavcodec */
		avcodec_init();
		avcodec_register_all();
		avcodec_initialized=1;
	}

	if (ff_mjpeg_encode_init(j->s) < 0) {
		av_free(j->s);
		av_free(j);
		return NULL;
	}

	/* alloc bogus avctx to keep MPV_common_init from segfaulting */
	j->s->avctx = calloc(sizeof(*j->s->avctx), 1);
	/* Set up to encode mjpeg */
	j->s->avctx->codec_id = CODEC_ID_MJPEG;

	/* make MPV_common_init allocate important buffers, like s->block */
	j->s->avctx->thread_count = 1;

	if (MPV_common_init(j->s) < 0) {
		av_free(j->s);
		av_free(j);
		return NULL;
	}

	/* correct the value for sc->mb_height */
	j->s->mb_height = j->s->height/8;
	j->s->mb_intra = 1;

	j->s->intra_matrix[0] = ff_mpeg1_default_intra_matrix[0];
	for (i = 1; i < 64; i++) 
		j->s->intra_matrix[i] = av_clip_uint8(
			(ff_mpeg1_default_intra_matrix[i]*j->s->qscale) >> 3);
	convert_matrix(j->s, j->s->q_intra_matrix, j->s->q_intra_matrix16, 
			j->s->intra_matrix, j->s->intra_quant_bias, 8, 8);
	return j;
}	

int jpeg_enc_frame(jpeg_enc_t *j, unsigned char *y_data, 
		unsigned char *u_data, unsigned char *v_data, char *bufr) {
	int i, k, mb_x, mb_y, overflow;
	short int *dest;
	unsigned char *source;
	/* initialize the buffer */

	init_put_bits(&j->s->pb, bufr, 1024*256);

	ff_mjpeg_encode_picture_header(j->s);

	j->s->header_bits = put_bits_count(&j->s->pb);

	j->s->last_dc[0] = 128; 
	j->s->last_dc[1] = 128; 
	j->s->last_dc[2] = 128;

	for (mb_y = 0; mb_y < j->s->mb_height; mb_y++) {
		for (mb_x = 0; mb_x < j->s->mb_width; mb_x++) {
			/* conversion 8 to 16 bit and filling of blocks
			 * must be mmx optimized */
			/* fill 2 Y macroblocks and one U and one V */
			source = mb_y * 8 * j->y_rs + 
				16 * j->y_ps * mb_x + y_data;
			dest = j->s->block[0];
			for (i = 0; i < 8; i++) {
				for (k = 0; k < 8; k++) {
					dest[k] = source[k*j->y_ps];
				}
				dest += 8;
				source += j->y_rs;
			}
			source = mb_y * 8 * j->y_rs + 
				(16*mb_x + 8)*j->y_ps + y_data;
			dest = j->s->block[1];
			for (i = 0; i < 8; i++) {
				for (k = 0; k < 8; k++) {
					dest[k] = source[k*j->y_ps];
				}
				dest += 8;
				source += j->y_rs;
			}
			if (!j->bw && j->cheap_upsample) {
				source = mb_y*4*j->u_rs + 
					8*mb_x*j->u_ps + u_data;
				dest = j->s->block[2];
				for (i = 0; i < 4; i++) {
					for (k = 0; k < 8; k++) {
						dest[k] = source[k*j->u_ps];
						dest[k+8] = source[k*j->u_ps];
					}
					dest += 16;
					source += j->u_rs;
				}
				source = mb_y*4*j->v_rs + 
					8*mb_x*j->v_ps + v_data;
				dest = j->s->block[3];
				for (i = 0; i < 4; i++) {
					for (k = 0; k < 8; k++) {
						dest[k] = source[k*j->v_ps];
						dest[k+8] = source[k*j->v_ps];
					}
					dest += 16;
					source += j->u_rs;
				}
			} else if (!j->bw && !j->cheap_upsample) {
				source = mb_y*8*j->u_rs + 
					8*mb_x*j->u_ps + u_data;
				dest = j->s->block[2];
				for (i = 0; i < 8; i++) {
					for (k = 0; k < 8; k++) 
						dest[k] = source[k*j->u_ps];
					dest += 8;
					source += j->u_rs;
				}
				source = mb_y*8*j->v_rs + 
					8*mb_x*j->v_ps + v_data;
				dest = j->s->block[3];
				for (i = 0; i < 8; i++) {
					for (k = 0; k < 8; k++) 
						dest[k] = source[k*j->v_ps];
					dest += 8;
					source += j->u_rs;
				}
			}
			emms_c(); /* is this really needed? */

			j->s->block_last_index[0] = 
				j->s->dct_quantize(j->s, j->s->block[0], 
						0, 8, &overflow);
			if (overflow) clip_coeffs(j->s, j->s->block[0], 
					j->s->block_last_index[0]);
			j->s->block_last_index[1] = 
				j->s->dct_quantize(j->s, j->s->block[1], 
						1, 8, &overflow);
			if (overflow) clip_coeffs(j->s, j->s->block[1], 
					j->s->block_last_index[1]);

			if (!j->bw) {
				j->s->block_last_index[4] =
					j->s->dct_quantize(j->s, j->s->block[2],
							4, 8, &overflow);
				if (overflow) clip_coeffs(j->s, j->s->block[2], 
						j->s->block_last_index[2]);
				j->s->block_last_index[5] =
					j->s->dct_quantize(j->s, j->s->block[3],
							5, 8, &overflow);
				if (overflow) clip_coeffs(j->s, j->s->block[3], 
						j->s->block_last_index[3]);
			}
			zr_mjpeg_encode_mb(j);
		}
	}
	emms_c();
	ff_mjpeg_encode_picture_trailer(j->s);
	flush_put_bits(&j->s->pb);	

	//FIXME
	//if (j->s->mjpeg_write_tables == 1)
	//	j->s->mjpeg_write_tables = 0;
	
	return pbBufPtr(&(j->s->pb)) - j->s->pb.buf;
}

void jpeg_enc_uninit(jpeg_enc_t *j) {
	ff_mjpeg_encode_close(j->s);
	av_free(j->s);
	av_free(j);
}