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 {
63 static void XAVS_log(void *p, int level, const char *fmt, va_list args)
65 static const int level_map[] = {
66 [XAVS_LOG_ERROR] = AV_LOG_ERROR,
67 [XAVS_LOG_WARNING] = AV_LOG_WARNING,
68 [XAVS_LOG_INFO] = AV_LOG_INFO,
69 [XAVS_LOG_DEBUG] = AV_LOG_DEBUG
72 if (level < 0 || level > XAVS_LOG_DEBUG)
75 av_vlog(p, level_map[level], fmt, args);
78 static int encode_nals(AVCodecContext *ctx, AVPacket *pkt,
79 xavs_nal_t *nals, int nnal)
81 XavsContext *x4 = ctx->priv_data;
83 int i, s, ret, size = x4->sei_size + FF_MIN_BUFFER_SIZE;
88 for (i = 0; i < nnal; i++)
89 size += nals[i].i_payload;
91 if ((ret = ff_alloc_packet2(ctx, pkt, size)) < 0)
95 /* Write the SEI as part of the first frame. */
96 if (x4->sei_size > 0 && nnal > 0) {
97 memcpy(p, x4->sei, x4->sei_size);
102 for (i = 0; i < nnal; i++) {
103 s = xavs_nal_encode(p, &size, 1, nals + i);
108 pkt->size = p - pkt->data;
113 static int XAVS_frame(AVCodecContext *avctx, AVPacket *pkt,
114 const AVFrame *frame, int *got_packet)
116 XavsContext *x4 = avctx->priv_data;
119 xavs_picture_t pic_out;
121 x4->pic.img.i_csp = XAVS_CSP_I420;
122 x4->pic.img.i_plane = 3;
125 for (i = 0; i < 3; i++) {
126 x4->pic.img.plane[i] = frame->data[i];
127 x4->pic.img.i_stride[i] = frame->linesize[i];
130 x4->pic.i_pts = frame->pts;
131 x4->pic.i_type = XAVS_TYPE_AUTO;
132 x4->pts_buffer[avctx->frame_number % (avctx->max_b_frames+1)] = frame->pts;
135 if (xavs_encoder_encode(x4->enc, &nal, &nnal,
136 frame? &x4->pic: NULL, &pic_out) < 0)
139 ret = encode_nals(avctx, pkt, nal, nnal);
145 if (!frame && !(x4->end_of_stream)) {
146 if ((ret = ff_alloc_packet2(avctx, pkt, 4)) < 0)
153 pkt->dts = 2*x4->pts_buffer[(x4->out_frame_count-1)%(avctx->max_b_frames+1)] -
154 x4->pts_buffer[(x4->out_frame_count-2)%(avctx->max_b_frames+1)];
155 x4->end_of_stream = END_OF_STREAM;
161 avctx->coded_frame->pts = pic_out.i_pts;
162 pkt->pts = pic_out.i_pts;
163 if (avctx->has_b_frames) {
164 if (!x4->out_frame_count)
165 pkt->dts = pkt->pts - (x4->pts_buffer[1] - x4->pts_buffer[0]);
167 pkt->dts = x4->pts_buffer[(x4->out_frame_count-1)%(avctx->max_b_frames+1)];
171 switch (pic_out.i_type) {
174 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_I;
177 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_P;
181 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_B;
185 /* There is no IDR frame in AVS JiZhun */
186 /* Sequence header is used as a flag */
187 if (pic_out.i_type == XAVS_TYPE_I) {
188 avctx->coded_frame->key_frame = 1;
189 pkt->flags |= AV_PKT_FLAG_KEY;
192 avctx->coded_frame->quality = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
194 x4->out_frame_count++;
199 static av_cold int XAVS_close(AVCodecContext *avctx)
201 XavsContext *x4 = avctx->priv_data;
203 av_freep(&avctx->extradata);
205 av_freep(&x4->pts_buffer);
208 xavs_encoder_close(x4->enc);
210 av_frame_free(&avctx->coded_frame);
215 static av_cold int XAVS_init(AVCodecContext *avctx)
217 XavsContext *x4 = avctx->priv_data;
220 xavs_param_default(&x4->params);
222 x4->params.pf_log = XAVS_log;
223 x4->params.p_log_private = avctx;
224 x4->params.i_keyint_max = avctx->gop_size;
225 if (avctx->bit_rate) {
226 x4->params.rc.i_bitrate = avctx->bit_rate / 1000;
227 x4->params.rc.i_rc_method = XAVS_RC_ABR;
229 x4->params.rc.i_vbv_buffer_size = avctx->rc_buffer_size / 1000;
230 x4->params.rc.i_vbv_max_bitrate = avctx->rc_max_rate / 1000;
231 x4->params.rc.b_stat_write = avctx->flags & CODEC_FLAG_PASS1;
232 if (avctx->flags & CODEC_FLAG_PASS2) {
233 x4->params.rc.b_stat_read = 1;
236 x4->params.rc.i_rc_method = XAVS_RC_CRF;
237 x4->params.rc.f_rf_constant = x4->crf;
238 } else if (x4->cqp >= 0) {
239 x4->params.rc.i_rc_method = XAVS_RC_CQP;
240 x4->params.rc.i_qp_constant = x4->cqp;
245 x4->params.b_aud = x4->aud;
247 x4->params.rc.b_mb_tree = x4->mbtree;
248 if (x4->direct_pred >= 0)
249 x4->params.analyse.i_direct_mv_pred = x4->direct_pred;
250 if (x4->fast_pskip >= 0)
251 x4->params.analyse.b_fast_pskip = x4->fast_pskip;
252 if (x4->mixed_refs >= 0)
253 x4->params.analyse.b_mixed_references = x4->mixed_refs;
254 if (x4->b_bias != INT_MIN)
255 x4->params.i_bframe_bias = x4->b_bias;
256 if (x4->cplxblur >= 0)
257 x4->params.rc.f_complexity_blur = x4->cplxblur;
259 x4->params.i_bframe = avctx->max_b_frames;
260 /* cabac is not included in AVS JiZhun Profile */
261 x4->params.b_cabac = 0;
263 x4->params.i_bframe_adaptive = avctx->b_frame_strategy;
265 avctx->has_b_frames = !!avctx->max_b_frames;
267 /* AVS doesn't allow B picture as reference */
268 /* The max allowed reference frame number of B is 2 */
269 x4->params.i_keyint_min = avctx->keyint_min;
270 if (x4->params.i_keyint_min > x4->params.i_keyint_max)
271 x4->params.i_keyint_min = x4->params.i_keyint_max;
273 x4->params.i_scenecut_threshold = avctx->scenechange_threshold;
275 // x4->params.b_deblocking_filter = avctx->flags & CODEC_FLAG_LOOP_FILTER;
277 x4->params.rc.i_qp_min = avctx->qmin;
278 x4->params.rc.i_qp_max = avctx->qmax;
279 x4->params.rc.i_qp_step = avctx->max_qdiff;
281 x4->params.rc.f_qcompress = avctx->qcompress; /* 0.0 => cbr, 1.0 => constant qp */
282 x4->params.rc.f_qblur = avctx->qblur; /* temporally blur quants */
284 x4->params.i_frame_reference = avctx->refs;
286 x4->params.i_width = avctx->width;
287 x4->params.i_height = avctx->height;
288 x4->params.vui.i_sar_width = avctx->sample_aspect_ratio.num;
289 x4->params.vui.i_sar_height = avctx->sample_aspect_ratio.den;
290 /* This is only used for counting the fps */
291 x4->params.i_fps_num = avctx->time_base.den;
292 x4->params.i_fps_den = avctx->time_base.num;
293 x4->params.analyse.inter = XAVS_ANALYSE_I8x8 |XAVS_ANALYSE_PSUB16x16| XAVS_ANALYSE_BSUB16x16;
295 switch (avctx->me_method) {
297 x4->params.analyse.i_me_method = XAVS_ME_DIA;
300 x4->params.analyse.i_me_method = XAVS_ME_HEX;
303 x4->params.analyse.i_me_method = XAVS_ME_UMH;
306 x4->params.analyse.i_me_method = XAVS_ME_ESA;
309 x4->params.analyse.i_me_method = XAVS_ME_TESA;
312 x4->params.analyse.i_me_method = XAVS_ME_HEX;
315 x4->params.analyse.i_me_range = avctx->me_range;
316 x4->params.analyse.i_subpel_refine = avctx->me_subpel_quality;
318 x4->params.analyse.b_chroma_me = avctx->me_cmp & FF_CMP_CHROMA;
319 /* AVS P2 only enables 8x8 transform */
320 x4->params.analyse.b_transform_8x8 = 1; //avctx->flags2 & CODEC_FLAG2_8X8DCT;
322 x4->params.analyse.i_trellis = avctx->trellis;
323 x4->params.analyse.i_noise_reduction = avctx->noise_reduction;
325 if (avctx->level > 0)
326 x4->params.i_level_idc = avctx->level;
328 x4->params.rc.f_rate_tolerance =
329 (float)avctx->bit_rate_tolerance/avctx->bit_rate;
331 if ((avctx->rc_buffer_size) &&
332 (avctx->rc_initial_buffer_occupancy <= avctx->rc_buffer_size)) {
333 x4->params.rc.f_vbv_buffer_init =
334 (float)avctx->rc_initial_buffer_occupancy / avctx->rc_buffer_size;
336 x4->params.rc.f_vbv_buffer_init = 0.9;
338 /* TAG:do we have MB tree RC method */
339 /* what is the RC method we are now using? Default NO */
340 x4->params.rc.f_ip_factor = 1 / fabs(avctx->i_quant_factor);
341 x4->params.rc.f_pb_factor = avctx->b_quant_factor;
342 x4->params.analyse.i_chroma_qp_offset = avctx->chromaoffset;
344 x4->params.analyse.b_psnr = avctx->flags & CODEC_FLAG_PSNR;
345 x4->params.i_log_level = XAVS_LOG_DEBUG;
346 x4->params.i_threads = avctx->thread_count;
347 x4->params.b_interlaced = avctx->flags & CODEC_FLAG_INTERLACED_DCT;
349 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER)
350 x4->params.b_repeat_headers = 0;
352 x4->enc = xavs_encoder_open(&x4->params);
356 if (!(x4->pts_buffer = av_mallocz_array((avctx->max_b_frames+1), sizeof(*x4->pts_buffer))))
357 return AVERROR(ENOMEM);
359 avctx->coded_frame = av_frame_alloc();
360 if (!avctx->coded_frame)
361 return AVERROR(ENOMEM);
363 /* TAG: Do we have GLOBAL HEADER in AVS */
364 /* We Have PPS and SPS in AVS */
365 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER && 0) {
367 int nnal, s, i, size;
370 s = xavs_encoder_headers(x4->enc, &nal, &nnal);
372 avctx->extradata = p = av_malloc(s);
373 for (i = 0; i < nnal; i++) {
374 /* Don't put the SEI in extradata. */
375 if (nal[i].i_type == NAL_SEI) {
376 x4->sei = av_malloc( 5 + nal[i].i_payload * 4 / 3 );
377 if (xavs_nal_encode(x4->sei, &x4->sei_size, 1, nal + i) < 0)
382 size = xavs_nal_encode(p, &s, 1, nal + i);
387 avctx->extradata_size = p - avctx->extradata;
392 #define OFFSET(x) offsetof(XavsContext, x)
393 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
394 static const AVOption options[] = {
395 { "crf", "Select the quality for constant quality mode", OFFSET(crf), AV_OPT_TYPE_FLOAT, {.dbl = -1 }, -1, FLT_MAX, VE },
396 { "qp", "Constant quantization parameter rate control method",OFFSET(cqp), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, INT_MAX, VE },
397 { "b-bias", "Influences how often B-frames are used", OFFSET(b_bias), AV_OPT_TYPE_INT, {.i64 = INT_MIN}, INT_MIN, INT_MAX, VE },
398 { "cplxblur", "Reduce fluctuations in QP (before curve compression)", OFFSET(cplxblur), AV_OPT_TYPE_FLOAT, {.dbl = -1 }, -1, FLT_MAX, VE},
399 { "direct-pred", "Direct MV prediction mode", OFFSET(direct_pred), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, INT_MAX, VE, "direct-pred" },
400 { "none", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_NONE }, 0, 0, VE, "direct-pred" },
401 { "spatial", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_SPATIAL }, 0, 0, VE, "direct-pred" },
402 { "temporal", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_TEMPORAL }, 0, 0, VE, "direct-pred" },
403 { "auto", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_AUTO }, 0, 0, VE, "direct-pred" },
404 { "aud", "Use access unit delimiters.", OFFSET(aud), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, 1, VE},
405 { "mbtree", "Use macroblock tree ratecontrol.", OFFSET(mbtree), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, 1, VE},
406 { "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 },
407 { "fast-pskip", NULL, OFFSET(fast_pskip), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, 1, VE},
411 static const AVClass xavs_class = {
412 .class_name = "libxavs",
413 .item_name = av_default_item_name,
415 .version = LIBAVUTIL_VERSION_INT,
418 static const AVCodecDefault xavs_defaults[] = {
423 AVCodec ff_libxavs_encoder = {
425 .long_name = NULL_IF_CONFIG_SMALL("libxavs Chinese AVS (Audio Video Standard)"),
426 .type = AVMEDIA_TYPE_VIDEO,
427 .id = AV_CODEC_ID_CAVS,
428 .priv_data_size = sizeof(XavsContext),
430 .encode2 = XAVS_frame,
432 .capabilities = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS,
433 .pix_fmts = (const enum AVPixelFormat[]) { AV_PIX_FMT_YUV420P, AV_PIX_FMT_NONE },
434 .priv_class = &xavs_class,
435 .defaults = xavs_defaults,