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 {
64 static void XAVS_log(void *p, int level, const char *fmt, va_list args)
66 static const int level_map[] = {
67 [XAVS_LOG_ERROR] = AV_LOG_ERROR,
68 [XAVS_LOG_WARNING] = AV_LOG_WARNING,
69 [XAVS_LOG_INFO] = AV_LOG_INFO,
70 [XAVS_LOG_DEBUG] = AV_LOG_DEBUG
73 if (level < 0 || level > XAVS_LOG_DEBUG)
76 av_vlog(p, level_map[level], fmt, args);
79 static int encode_nals(AVCodecContext *ctx, AVPacket *pkt,
80 xavs_nal_t *nals, int nnal)
82 XavsContext *x4 = ctx->priv_data;
84 int i, s, ret, size = x4->sei_size + AV_INPUT_BUFFER_MIN_SIZE;
89 for (i = 0; i < nnal; i++)
90 size += nals[i].i_payload;
92 if ((ret = ff_alloc_packet2(ctx, pkt, size, 0)) < 0)
96 /* Write the SEI as part of the first frame. */
97 if (x4->sei_size > 0 && nnal > 0) {
98 memcpy(p, x4->sei, x4->sei_size);
103 for (i = 0; i < nnal; i++) {
104 s = xavs_nal_encode(p, &size, 1, nals + i);
109 pkt->size = p - pkt->data;
114 static int XAVS_frame(AVCodecContext *avctx, AVPacket *pkt,
115 const AVFrame *frame, int *got_packet)
117 XavsContext *x4 = avctx->priv_data;
120 xavs_picture_t pic_out;
123 x4->pic.img.i_csp = XAVS_CSP_I420;
124 x4->pic.img.i_plane = 3;
127 for (i = 0; i < 3; i++) {
128 x4->pic.img.plane[i] = frame->data[i];
129 x4->pic.img.i_stride[i] = frame->linesize[i];
132 x4->pic.i_pts = frame->pts;
133 x4->pic.i_type = XAVS_TYPE_AUTO;
134 x4->pts_buffer[avctx->frame_number % (avctx->max_b_frames+1)] = frame->pts;
137 if (xavs_encoder_encode(x4->enc, &nal, &nnal,
138 frame? &x4->pic: NULL, &pic_out) < 0)
141 ret = encode_nals(avctx, pkt, nal, nnal);
147 if (!frame && !(x4->end_of_stream)) {
148 if ((ret = ff_alloc_packet2(avctx, pkt, 4, 0)) < 0)
155 pkt->dts = 2*x4->pts_buffer[(x4->out_frame_count-1)%(avctx->max_b_frames+1)] -
156 x4->pts_buffer[(x4->out_frame_count-2)%(avctx->max_b_frames+1)];
157 x4->end_of_stream = END_OF_STREAM;
163 #if FF_API_CODED_FRAME
164 FF_DISABLE_DEPRECATION_WARNINGS
165 avctx->coded_frame->pts = pic_out.i_pts;
166 FF_ENABLE_DEPRECATION_WARNINGS
168 pkt->pts = pic_out.i_pts;
169 if (avctx->has_b_frames) {
170 if (!x4->out_frame_count)
171 pkt->dts = pkt->pts - (x4->pts_buffer[1] - x4->pts_buffer[0]);
173 pkt->dts = x4->pts_buffer[(x4->out_frame_count-1)%(avctx->max_b_frames+1)];
177 switch (pic_out.i_type) {
180 pict_type = AV_PICTURE_TYPE_I;
183 pict_type = AV_PICTURE_TYPE_P;
187 pict_type = AV_PICTURE_TYPE_B;
190 pict_type = AV_PICTURE_TYPE_NONE;
192 #if FF_API_CODED_FRAME
193 FF_DISABLE_DEPRECATION_WARNINGS
194 avctx->coded_frame->pict_type = pict_type;
195 FF_ENABLE_DEPRECATION_WARNINGS
198 /* There is no IDR frame in AVS JiZhun */
199 /* Sequence header is used as a flag */
200 if (pic_out.i_type == XAVS_TYPE_I) {
201 #if FF_API_CODED_FRAME
202 FF_DISABLE_DEPRECATION_WARNINGS
203 avctx->coded_frame->key_frame = 1;
204 FF_ENABLE_DEPRECATION_WARNINGS
206 pkt->flags |= AV_PKT_FLAG_KEY;
209 #if FF_API_CODED_FRAME
210 FF_DISABLE_DEPRECATION_WARNINGS
211 avctx->coded_frame->quality = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
212 FF_ENABLE_DEPRECATION_WARNINGS
215 ff_side_data_set_encoder_stats(pkt, (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA, NULL, 0, pict_type);
217 x4->out_frame_count++;
222 static av_cold int XAVS_close(AVCodecContext *avctx)
224 XavsContext *x4 = avctx->priv_data;
226 av_freep(&avctx->extradata);
228 av_freep(&x4->pts_buffer);
231 xavs_encoder_close(x4->enc);
236 static av_cold int XAVS_init(AVCodecContext *avctx)
238 XavsContext *x4 = avctx->priv_data;
241 xavs_param_default(&x4->params);
243 x4->params.pf_log = XAVS_log;
244 x4->params.p_log_private = avctx;
245 x4->params.i_keyint_max = avctx->gop_size;
246 if (avctx->bit_rate) {
247 x4->params.rc.i_bitrate = avctx->bit_rate / 1000;
248 x4->params.rc.i_rc_method = XAVS_RC_ABR;
250 x4->params.rc.i_vbv_buffer_size = avctx->rc_buffer_size / 1000;
251 x4->params.rc.i_vbv_max_bitrate = avctx->rc_max_rate / 1000;
252 x4->params.rc.b_stat_write = avctx->flags & AV_CODEC_FLAG_PASS1;
253 if (avctx->flags & AV_CODEC_FLAG_PASS2) {
254 x4->params.rc.b_stat_read = 1;
257 x4->params.rc.i_rc_method = XAVS_RC_CRF;
258 x4->params.rc.f_rf_constant = x4->crf;
259 } else if (x4->cqp >= 0) {
260 x4->params.rc.i_rc_method = XAVS_RC_CQP;
261 x4->params.rc.i_qp_constant = x4->cqp;
266 x4->params.b_aud = x4->aud;
268 x4->params.rc.b_mb_tree = x4->mbtree;
269 if (x4->direct_pred >= 0)
270 x4->params.analyse.i_direct_mv_pred = x4->direct_pred;
271 if (x4->fast_pskip >= 0)
272 x4->params.analyse.b_fast_pskip = x4->fast_pskip;
273 if (x4->motion_est >= 0)
274 x4->params.analyse.i_me_method = x4->motion_est;
275 if (x4->mixed_refs >= 0)
276 x4->params.analyse.b_mixed_references = x4->mixed_refs;
277 if (x4->b_bias != INT_MIN)
278 x4->params.i_bframe_bias = x4->b_bias;
279 if (x4->cplxblur >= 0)
280 x4->params.rc.f_complexity_blur = x4->cplxblur;
282 #if FF_API_MOTION_EST
283 FF_DISABLE_DEPRECATION_WARNINGS
284 if (x4->motion_est < 0) {
285 switch (avctx->me_method) {
287 x4->params.analyse.i_me_method = XAVS_ME_DIA;
290 x4->params.analyse.i_me_method = XAVS_ME_HEX;
293 x4->params.analyse.i_me_method = XAVS_ME_UMH;
296 x4->params.analyse.i_me_method = XAVS_ME_ESA;
299 x4->params.analyse.i_me_method = XAVS_ME_TESA;
302 x4->params.analyse.i_me_method = XAVS_ME_HEX;
305 FF_ENABLE_DEPRECATION_WARNINGS
308 x4->params.i_bframe = avctx->max_b_frames;
309 /* cabac is not included in AVS JiZhun Profile */
310 x4->params.b_cabac = 0;
312 x4->params.i_bframe_adaptive = avctx->b_frame_strategy;
314 avctx->has_b_frames = !!avctx->max_b_frames;
316 /* AVS doesn't allow B picture as reference */
317 /* The max allowed reference frame number of B is 2 */
318 x4->params.i_keyint_min = avctx->keyint_min;
319 if (x4->params.i_keyint_min > x4->params.i_keyint_max)
320 x4->params.i_keyint_min = x4->params.i_keyint_max;
322 x4->params.i_scenecut_threshold = avctx->scenechange_threshold;
324 // x4->params.b_deblocking_filter = avctx->flags & AV_CODEC_FLAG_LOOP_FILTER;
326 x4->params.rc.i_qp_min = avctx->qmin;
327 x4->params.rc.i_qp_max = avctx->qmax;
328 x4->params.rc.i_qp_step = avctx->max_qdiff;
330 x4->params.rc.f_qcompress = avctx->qcompress; /* 0.0 => cbr, 1.0 => constant qp */
331 x4->params.rc.f_qblur = avctx->qblur; /* temporally blur quants */
333 x4->params.i_frame_reference = avctx->refs;
335 x4->params.i_width = avctx->width;
336 x4->params.i_height = avctx->height;
337 x4->params.vui.i_sar_width = avctx->sample_aspect_ratio.num;
338 x4->params.vui.i_sar_height = avctx->sample_aspect_ratio.den;
339 /* This is only used for counting the fps */
340 x4->params.i_fps_num = avctx->time_base.den;
341 x4->params.i_fps_den = avctx->time_base.num;
342 x4->params.analyse.inter = XAVS_ANALYSE_I8x8 |XAVS_ANALYSE_PSUB16x16| XAVS_ANALYSE_BSUB16x16;
344 x4->params.analyse.i_me_range = avctx->me_range;
345 x4->params.analyse.i_subpel_refine = avctx->me_subpel_quality;
347 x4->params.analyse.b_chroma_me = avctx->me_cmp & FF_CMP_CHROMA;
348 /* AVS P2 only enables 8x8 transform */
349 x4->params.analyse.b_transform_8x8 = 1; //avctx->flags2 & AV_CODEC_FLAG2_8X8DCT;
351 x4->params.analyse.i_trellis = avctx->trellis;
352 x4->params.analyse.i_noise_reduction = avctx->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;
372 x4->params.analyse.i_chroma_qp_offset = avctx->chromaoffset;
374 x4->params.analyse.b_psnr = avctx->flags & AV_CODEC_FLAG_PSNR;
375 x4->params.i_log_level = XAVS_LOG_DEBUG;
376 x4->params.i_threads = avctx->thread_count;
377 x4->params.b_interlaced = avctx->flags & AV_CODEC_FLAG_INTERLACED_DCT;
379 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER)
380 x4->params.b_repeat_headers = 0;
382 x4->enc = xavs_encoder_open(&x4->params);
386 if (!(x4->pts_buffer = av_mallocz_array((avctx->max_b_frames+1), sizeof(*x4->pts_buffer))))
387 return AVERROR(ENOMEM);
389 /* TAG: Do we have GLOBAL HEADER in AVS */
390 /* We Have PPS and SPS in AVS */
391 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER && 0) {
393 int nnal, s, i, size;
396 s = xavs_encoder_headers(x4->enc, &nal, &nnal);
398 avctx->extradata = p = av_malloc(s);
399 for (i = 0; i < nnal; i++) {
400 /* Don't put the SEI in extradata. */
401 if (nal[i].i_type == NAL_SEI) {
402 x4->sei = av_malloc( 5 + nal[i].i_payload * 4 / 3 );
403 if (xavs_nal_encode(x4->sei, &x4->sei_size, 1, nal + i) < 0)
408 size = xavs_nal_encode(p, &s, 1, nal + i);
413 avctx->extradata_size = p - avctx->extradata;
418 #define OFFSET(x) offsetof(XavsContext, x)
419 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
420 static const AVOption options[] = {
421 { "crf", "Select the quality for constant quality mode", OFFSET(crf), AV_OPT_TYPE_FLOAT, {.dbl = -1 }, -1, FLT_MAX, VE },
422 { "qp", "Constant quantization parameter rate control method",OFFSET(cqp), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, INT_MAX, VE },
423 { "b-bias", "Influences how often B-frames are used", OFFSET(b_bias), AV_OPT_TYPE_INT, {.i64 = INT_MIN}, INT_MIN, INT_MAX, VE },
424 { "cplxblur", "Reduce fluctuations in QP (before curve compression)", OFFSET(cplxblur), AV_OPT_TYPE_FLOAT, {.dbl = -1 }, -1, FLT_MAX, VE},
425 { "direct-pred", "Direct MV prediction mode", OFFSET(direct_pred), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, INT_MAX, VE, "direct-pred" },
426 { "none", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_NONE }, 0, 0, VE, "direct-pred" },
427 { "spatial", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_SPATIAL }, 0, 0, VE, "direct-pred" },
428 { "temporal", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_TEMPORAL }, 0, 0, VE, "direct-pred" },
429 { "auto", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_AUTO }, 0, 0, VE, "direct-pred" },
430 { "aud", "Use access unit delimiters.", OFFSET(aud), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, 1, VE},
431 { "mbtree", "Use macroblock tree ratecontrol.", OFFSET(mbtree), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, 1, VE},
432 { "mixed-refs", "One reference per partition, as opposed to one reference per macroblock", OFFSET(mixed_refs), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE },
433 { "fast-pskip", NULL, OFFSET(fast_pskip), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, 1, VE},
434 { "motion-est", "Set motion estimation method", OFFSET(motion_est), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, XAVS_ME_TESA, VE, "motion-est"},
435 { "dia", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_DIA }, INT_MIN, INT_MAX, VE, "motion-est" },
436 { "hex", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_HEX }, INT_MIN, INT_MAX, VE, "motion-est" },
437 { "umh", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_UMH }, INT_MIN, INT_MAX, VE, "motion-est" },
438 { "esa", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_ESA }, INT_MIN, INT_MAX, VE, "motion-est" },
439 { "tesa", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_TESA }, INT_MIN, INT_MAX, VE, "motion-est" },
444 static const AVClass xavs_class = {
445 .class_name = "libxavs",
446 .item_name = av_default_item_name,
448 .version = LIBAVUTIL_VERSION_INT,
451 static const AVCodecDefault xavs_defaults[] = {
456 AVCodec ff_libxavs_encoder = {
458 .long_name = NULL_IF_CONFIG_SMALL("libxavs Chinese AVS (Audio Video Standard)"),
459 .type = AVMEDIA_TYPE_VIDEO,
460 .id = AV_CODEC_ID_CAVS,
461 .priv_data_size = sizeof(XavsContext),
463 .encode2 = XAVS_frame,
465 .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS,
466 .pix_fmts = (const enum AVPixelFormat[]) { AV_PIX_FMT_YUV420P, AV_PIX_FMT_NONE },
467 .priv_class = &xavs_class,
468 .defaults = xavs_defaults,