2 * AVS encoding using the xavs library
3 * Copyright (C) 2010 Amanda, Y.N. Wu <amanda11192003@gmail.com>
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
31 #include "libavutil/internal.h"
32 #include "libavutil/mem.h"
33 #include "libavutil/opt.h"
35 #define END_OF_STREAM 0x001
37 #define XAVS_PART_I8X8 0x002 /* Analyze i8x8 (requires 8x8 transform) */
38 #define XAVS_PART_P8X8 0x010 /* Analyze p16x8, p8x16 and p8x8 */
39 #define XAVS_PART_B8X8 0x100 /* Analyze b16x8, b*/
41 typedef struct XavsContext {
61 int scenechange_threshold;
68 static void XAVS_log(void *p, int level, const char *fmt, va_list args)
70 static const int level_map[] = {
71 [XAVS_LOG_ERROR] = AV_LOG_ERROR,
72 [XAVS_LOG_WARNING] = AV_LOG_WARNING,
73 [XAVS_LOG_INFO] = AV_LOG_INFO,
74 [XAVS_LOG_DEBUG] = AV_LOG_DEBUG
77 if (level < 0 || level > XAVS_LOG_DEBUG)
80 av_vlog(p, level_map[level], fmt, args);
83 static int encode_nals(AVCodecContext *ctx, AVPacket *pkt,
84 xavs_nal_t *nals, int nnal)
86 XavsContext *x4 = ctx->priv_data;
88 int i, s, ret, size = x4->sei_size + AV_INPUT_BUFFER_MIN_SIZE;
93 for (i = 0; i < nnal; i++)
94 size += nals[i].i_payload;
96 if ((ret = ff_alloc_packet2(ctx, pkt, size, 0)) < 0)
100 /* Write the SEI as part of the first frame. */
101 if (x4->sei_size > 0 && nnal > 0) {
102 memcpy(p, x4->sei, x4->sei_size);
107 for (i = 0; i < nnal; i++) {
108 s = xavs_nal_encode(p, &size, 1, nals + i);
113 pkt->size = p - pkt->data;
118 static int XAVS_frame(AVCodecContext *avctx, AVPacket *pkt,
119 const AVFrame *frame, int *got_packet)
121 XavsContext *x4 = avctx->priv_data;
124 xavs_picture_t pic_out;
127 x4->pic.img.i_csp = XAVS_CSP_I420;
128 x4->pic.img.i_plane = 3;
131 for (i = 0; i < 3; i++) {
132 x4->pic.img.plane[i] = frame->data[i];
133 x4->pic.img.i_stride[i] = frame->linesize[i];
136 x4->pic.i_pts = frame->pts;
137 x4->pic.i_type = XAVS_TYPE_AUTO;
138 x4->pts_buffer[avctx->frame_number % (avctx->max_b_frames+1)] = frame->pts;
141 if (xavs_encoder_encode(x4->enc, &nal, &nnal,
142 frame? &x4->pic: NULL, &pic_out) < 0)
145 ret = encode_nals(avctx, pkt, nal, nnal);
151 if (!frame && !(x4->end_of_stream)) {
152 if ((ret = ff_alloc_packet2(avctx, pkt, 4, 0)) < 0)
159 pkt->dts = 2*x4->pts_buffer[(x4->out_frame_count-1)%(avctx->max_b_frames+1)] -
160 x4->pts_buffer[(x4->out_frame_count-2)%(avctx->max_b_frames+1)];
161 x4->end_of_stream = END_OF_STREAM;
167 #if FF_API_CODED_FRAME
168 FF_DISABLE_DEPRECATION_WARNINGS
169 avctx->coded_frame->pts = pic_out.i_pts;
170 FF_ENABLE_DEPRECATION_WARNINGS
172 pkt->pts = pic_out.i_pts;
173 if (avctx->has_b_frames) {
174 if (!x4->out_frame_count)
175 pkt->dts = pkt->pts - (x4->pts_buffer[1] - x4->pts_buffer[0]);
177 pkt->dts = x4->pts_buffer[(x4->out_frame_count-1)%(avctx->max_b_frames+1)];
181 switch (pic_out.i_type) {
184 pict_type = AV_PICTURE_TYPE_I;
187 pict_type = AV_PICTURE_TYPE_P;
191 pict_type = AV_PICTURE_TYPE_B;
194 pict_type = AV_PICTURE_TYPE_NONE;
196 #if FF_API_CODED_FRAME
197 FF_DISABLE_DEPRECATION_WARNINGS
198 avctx->coded_frame->pict_type = pict_type;
199 FF_ENABLE_DEPRECATION_WARNINGS
202 /* There is no IDR frame in AVS JiZhun */
203 /* Sequence header is used as a flag */
204 if (pic_out.i_type == XAVS_TYPE_I) {
205 #if FF_API_CODED_FRAME
206 FF_DISABLE_DEPRECATION_WARNINGS
207 avctx->coded_frame->key_frame = 1;
208 FF_ENABLE_DEPRECATION_WARNINGS
210 pkt->flags |= AV_PKT_FLAG_KEY;
213 #if FF_API_CODED_FRAME
214 FF_DISABLE_DEPRECATION_WARNINGS
215 avctx->coded_frame->quality = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
216 FF_ENABLE_DEPRECATION_WARNINGS
219 ff_side_data_set_encoder_stats(pkt, (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA, NULL, 0, pict_type);
221 x4->out_frame_count++;
226 static av_cold int XAVS_close(AVCodecContext *avctx)
228 XavsContext *x4 = avctx->priv_data;
230 av_freep(&avctx->extradata);
232 av_freep(&x4->pts_buffer);
235 xavs_encoder_close(x4->enc);
240 static av_cold int XAVS_init(AVCodecContext *avctx)
242 XavsContext *x4 = avctx->priv_data;
245 xavs_param_default(&x4->params);
247 x4->params.pf_log = XAVS_log;
248 x4->params.p_log_private = avctx;
249 x4->params.i_keyint_max = avctx->gop_size;
250 if (avctx->bit_rate) {
251 x4->params.rc.i_bitrate = avctx->bit_rate / 1000;
252 x4->params.rc.i_rc_method = XAVS_RC_ABR;
254 x4->params.rc.i_vbv_buffer_size = avctx->rc_buffer_size / 1000;
255 x4->params.rc.i_vbv_max_bitrate = avctx->rc_max_rate / 1000;
256 x4->params.rc.b_stat_write = avctx->flags & AV_CODEC_FLAG_PASS1;
257 if (avctx->flags & AV_CODEC_FLAG_PASS2) {
258 x4->params.rc.b_stat_read = 1;
261 x4->params.rc.i_rc_method = XAVS_RC_CRF;
262 x4->params.rc.f_rf_constant = x4->crf;
263 } else if (x4->cqp >= 0) {
264 x4->params.rc.i_rc_method = XAVS_RC_CQP;
265 x4->params.rc.i_qp_constant = x4->cqp;
270 x4->params.b_aud = x4->aud;
272 x4->params.rc.b_mb_tree = x4->mbtree;
273 if (x4->direct_pred >= 0)
274 x4->params.analyse.i_direct_mv_pred = x4->direct_pred;
275 if (x4->fast_pskip >= 0)
276 x4->params.analyse.b_fast_pskip = x4->fast_pskip;
277 if (x4->motion_est >= 0)
278 x4->params.analyse.i_me_method = x4->motion_est;
279 if (x4->mixed_refs >= 0)
280 x4->params.analyse.b_mixed_references = x4->mixed_refs;
281 if (x4->b_bias != INT_MIN)
282 x4->params.i_bframe_bias = x4->b_bias;
283 if (x4->cplxblur >= 0)
284 x4->params.rc.f_complexity_blur = x4->cplxblur;
286 x4->params.i_bframe = avctx->max_b_frames;
287 /* cabac is not included in AVS JiZhun Profile */
288 x4->params.b_cabac = 0;
290 #if FF_API_PRIVATE_OPT
291 FF_DISABLE_DEPRECATION_WARNINGS
292 if (avctx->b_frame_strategy)
293 x4->b_frame_strategy = avctx->b_frame_strategy;
294 FF_ENABLE_DEPRECATION_WARNINGS
297 x4->params.i_bframe_adaptive = x4->b_frame_strategy;
299 avctx->has_b_frames = !!avctx->max_b_frames;
301 /* AVS doesn't allow B picture as reference */
302 /* The max allowed reference frame number of B is 2 */
303 x4->params.i_keyint_min = avctx->keyint_min;
304 if (x4->params.i_keyint_min > x4->params.i_keyint_max)
305 x4->params.i_keyint_min = x4->params.i_keyint_max;
307 #if FF_API_PRIVATE_OPT
308 FF_DISABLE_DEPRECATION_WARNINGS
309 if (avctx->scenechange_threshold)
310 x4->scenechange_threshold = avctx->scenechange_threshold;
311 FF_ENABLE_DEPRECATION_WARNINGS
314 x4->params.i_scenecut_threshold = x4->scenechange_threshold;
316 // x4->params.b_deblocking_filter = avctx->flags & AV_CODEC_FLAG_LOOP_FILTER;
318 x4->params.rc.i_qp_min = avctx->qmin;
319 x4->params.rc.i_qp_max = avctx->qmax;
320 x4->params.rc.i_qp_step = avctx->max_qdiff;
322 x4->params.rc.f_qcompress = avctx->qcompress; /* 0.0 => cbr, 1.0 => constant qp */
323 x4->params.rc.f_qblur = avctx->qblur; /* temporally blur quants */
325 x4->params.i_frame_reference = avctx->refs;
327 x4->params.i_width = avctx->width;
328 x4->params.i_height = avctx->height;
329 x4->params.vui.i_sar_width = avctx->sample_aspect_ratio.num;
330 x4->params.vui.i_sar_height = avctx->sample_aspect_ratio.den;
331 /* This is only used for counting the fps */
332 x4->params.i_fps_num = avctx->time_base.den;
333 x4->params.i_fps_den = avctx->time_base.num;
334 x4->params.analyse.inter = XAVS_ANALYSE_I8x8 |XAVS_ANALYSE_PSUB16x16| XAVS_ANALYSE_BSUB16x16;
336 x4->params.analyse.i_me_range = avctx->me_range;
337 x4->params.analyse.i_subpel_refine = avctx->me_subpel_quality;
339 x4->params.analyse.b_chroma_me = avctx->me_cmp & FF_CMP_CHROMA;
340 /* AVS P2 only enables 8x8 transform */
341 x4->params.analyse.b_transform_8x8 = 1; //avctx->flags2 & AV_CODEC_FLAG2_8X8DCT;
343 x4->params.analyse.i_trellis = avctx->trellis;
345 #if FF_API_PRIVATE_OPT
346 FF_DISABLE_DEPRECATION_WARNINGS
347 if (avctx->noise_reduction >= 0)
348 x4->noise_reduction = avctx->noise_reduction;
349 FF_ENABLE_DEPRECATION_WARNINGS
352 x4->params.analyse.i_noise_reduction = x4->noise_reduction;
354 if (avctx->level > 0)
355 x4->params.i_level_idc = avctx->level;
357 if (avctx->bit_rate > 0)
358 x4->params.rc.f_rate_tolerance =
359 (float)avctx->bit_rate_tolerance / avctx->bit_rate;
361 if ((avctx->rc_buffer_size) &&
362 (avctx->rc_initial_buffer_occupancy <= avctx->rc_buffer_size)) {
363 x4->params.rc.f_vbv_buffer_init =
364 (float)avctx->rc_initial_buffer_occupancy / avctx->rc_buffer_size;
366 x4->params.rc.f_vbv_buffer_init = 0.9;
368 /* TAG:do we have MB tree RC method */
369 /* what is the RC method we are now using? Default NO */
370 x4->params.rc.f_ip_factor = 1 / fabs(avctx->i_quant_factor);
371 x4->params.rc.f_pb_factor = avctx->b_quant_factor;
373 #if FF_API_PRIVATE_OPT
374 FF_DISABLE_DEPRECATION_WARNINGS
375 if (avctx->chromaoffset)
376 x4->chroma_offset = avctx->chromaoffset;
377 FF_ENABLE_DEPRECATION_WARNINGS
380 x4->params.analyse.i_chroma_qp_offset = x4->chroma_offset;
382 x4->params.analyse.b_psnr = avctx->flags & AV_CODEC_FLAG_PSNR;
383 x4->params.i_log_level = XAVS_LOG_DEBUG;
384 x4->params.i_threads = avctx->thread_count;
385 x4->params.b_interlaced = avctx->flags & AV_CODEC_FLAG_INTERLACED_DCT;
387 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER)
388 x4->params.b_repeat_headers = 0;
390 x4->enc = xavs_encoder_open(&x4->params);
394 if (!(x4->pts_buffer = av_mallocz_array((avctx->max_b_frames+1), sizeof(*x4->pts_buffer))))
395 return AVERROR(ENOMEM);
397 /* TAG: Do we have GLOBAL HEADER in AVS */
398 /* We Have PPS and SPS in AVS */
399 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER && 0) {
401 int nnal, s, i, size;
404 s = xavs_encoder_headers(x4->enc, &nal, &nnal);
406 avctx->extradata = p = av_malloc(s);
407 for (i = 0; i < nnal; i++) {
408 /* Don't put the SEI in extradata. */
409 if (nal[i].i_type == NAL_SEI) {
410 x4->sei = av_malloc( 5 + nal[i].i_payload * 4 / 3 );
411 if (xavs_nal_encode(x4->sei, &x4->sei_size, 1, nal + i) < 0)
416 size = xavs_nal_encode(p, &s, 1, nal + i);
421 avctx->extradata_size = p - avctx->extradata;
426 #define OFFSET(x) offsetof(XavsContext, x)
427 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
428 static const AVOption options[] = {
429 { "crf", "Select the quality for constant quality mode", OFFSET(crf), AV_OPT_TYPE_FLOAT, {.dbl = -1 }, -1, FLT_MAX, VE },
430 { "qp", "Constant quantization parameter rate control method",OFFSET(cqp), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, INT_MAX, VE },
431 { "b-bias", "Influences how often B-frames are used", OFFSET(b_bias), AV_OPT_TYPE_INT, {.i64 = INT_MIN}, INT_MIN, INT_MAX, VE },
432 { "cplxblur", "Reduce fluctuations in QP (before curve compression)", OFFSET(cplxblur), AV_OPT_TYPE_FLOAT, {.dbl = -1 }, -1, FLT_MAX, VE},
433 { "direct-pred", "Direct MV prediction mode", OFFSET(direct_pred), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, INT_MAX, VE, "direct-pred" },
434 { "none", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_NONE }, 0, 0, VE, "direct-pred" },
435 { "spatial", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_SPATIAL }, 0, 0, VE, "direct-pred" },
436 { "temporal", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_TEMPORAL }, 0, 0, VE, "direct-pred" },
437 { "auto", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_AUTO }, 0, 0, VE, "direct-pred" },
438 { "aud", "Use access unit delimiters.", OFFSET(aud), AV_OPT_TYPE_BOOL, {.i64 = -1 }, -1, 1, VE},
439 { "mbtree", "Use macroblock tree ratecontrol.", OFFSET(mbtree), AV_OPT_TYPE_BOOL, {.i64 = -1 }, -1, 1, VE},
440 { "mixed-refs", "One reference per partition, as opposed to one reference per macroblock", OFFSET(mixed_refs), AV_OPT_TYPE_BOOL, {.i64 = -1}, -1, 1, VE },
441 { "fast-pskip", NULL, OFFSET(fast_pskip), AV_OPT_TYPE_BOOL, {.i64 = -1 }, -1, 1, VE},
442 { "motion-est", "Set motion estimation method", OFFSET(motion_est), AV_OPT_TYPE_INT, { .i64 = XAVS_ME_DIA }, -1, XAVS_ME_TESA, VE, "motion-est"},
443 { "dia", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_DIA }, INT_MIN, INT_MAX, VE, "motion-est" },
444 { "hex", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_HEX }, INT_MIN, INT_MAX, VE, "motion-est" },
445 { "umh", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_UMH }, INT_MIN, INT_MAX, VE, "motion-est" },
446 { "esa", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_ESA }, INT_MIN, INT_MAX, VE, "motion-est" },
447 { "tesa", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_TESA }, INT_MIN, INT_MAX, VE, "motion-est" },
448 { "b_strategy", "Strategy to choose between I/P/B-frames", OFFSET(b_frame_strategy), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, 2, VE},
449 { "chromaoffset", "QP difference between chroma and luma", OFFSET(chroma_offset), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, VE},
450 { "sc_threshold", "Scene change threshold", OFFSET(scenechange_threshold), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, INT_MAX, VE},
451 { "noise_reduction", "Noise reduction", OFFSET(noise_reduction), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, INT_MAX, VE},
456 static const AVClass xavs_class = {
457 .class_name = "libxavs",
458 .item_name = av_default_item_name,
460 .version = LIBAVUTIL_VERSION_INT,
463 static const AVCodecDefault xavs_defaults[] = {
468 AVCodec ff_libxavs_encoder = {
470 .long_name = NULL_IF_CONFIG_SMALL("libxavs Chinese AVS (Audio Video Standard)"),
471 .type = AVMEDIA_TYPE_VIDEO,
472 .id = AV_CODEC_ID_CAVS,
473 .priv_data_size = sizeof(XavsContext),
475 .encode2 = XAVS_frame,
477 .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS,
478 .pix_fmts = (const enum AVPixelFormat[]) { AV_PIX_FMT_YUV420P, AV_PIX_FMT_NONE },
479 .priv_class = &xavs_class,
480 .defaults = xavs_defaults,
481 .wrapper_name = "libxavs",