2 * OpenH264 video encoder
3 * Copyright (C) 2014 Martin Storsjo
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
22 #include <wels/codec_api.h>
23 #include <wels/codec_ver.h>
25 #include "libavutil/attributes.h"
26 #include "libavutil/common.h"
27 #include "libavutil/opt.h"
28 #include "libavutil/internal.h"
29 #include "libavutil/intreadwrite.h"
30 #include "libavutil/mathematics.h"
34 #include "libopenh264.h"
36 #if !OPENH264_VER_AT_LEAST(1, 6)
37 #define SM_SIZELIMITED_SLICE SM_DYN_SLICE
40 #define TARGET_BITRATE_DEFAULT 2*1000*1000
42 typedef struct SVCContext {
43 const AVClass *av_class;
51 #if FF_API_OPENH264_CABAC
52 int cabac; // deprecated
60 #define OFFSET(x) offsetof(SVCContext, x)
61 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
62 #define DEPRECATED AV_OPT_FLAG_DEPRECATED
63 static const AVOption options[] = {
64 #if FF_API_OPENH264_SLICE_MODE
65 #if OPENH264_VER_AT_LEAST(1, 6)
66 { "slice_mode", "set slice mode, use slices/max_nal_size", OFFSET(slice_mode), AV_OPT_TYPE_INT, { .i64 = SM_FIXEDSLCNUM_SLICE }, SM_SINGLE_SLICE, SM_RESERVED, VE|DEPRECATED, "slice_mode" },
68 { "slice_mode", "set slice mode, use slices/max_nal_size", OFFSET(slice_mode), AV_OPT_TYPE_INT, { .i64 = SM_AUTO_SLICE }, SM_SINGLE_SLICE, SM_RESERVED, VE|DEPRECATED, "slice_mode" },
70 { "fixed", "a fixed number of slices", 0, AV_OPT_TYPE_CONST, { .i64 = SM_FIXEDSLCNUM_SLICE }, 0, 0, VE, "slice_mode" },
71 #if OPENH264_VER_AT_LEAST(1, 6)
72 { "dyn", "Size limited (compatibility name)", 0, AV_OPT_TYPE_CONST, { .i64 = SM_SIZELIMITED_SLICE }, 0, 0, VE, "slice_mode" },
73 { "sizelimited", "Size limited", 0, AV_OPT_TYPE_CONST, { .i64 = SM_SIZELIMITED_SLICE }, 0, 0, VE, "slice_mode" },
75 { "rowmb", "one slice per row of macroblocks", 0, AV_OPT_TYPE_CONST, { .i64 = SM_ROWMB_SLICE }, 0, 0, VE, "slice_mode" },
76 { "auto", "automatic number of slices according to number of threads", 0, AV_OPT_TYPE_CONST, { .i64 = SM_AUTO_SLICE }, 0, 0, VE, "slice_mode" },
77 { "dyn", "Dynamic slicing", 0, AV_OPT_TYPE_CONST, { .i64 = SM_DYN_SLICE }, 0, 0, VE, "slice_mode" },
80 { "loopfilter", "enable loop filter", OFFSET(loopfilter), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, VE },
81 { "profile", "set profile restrictions", OFFSET(profile), AV_OPT_TYPE_INT, { .i64 = FF_PROFILE_UNKNOWN }, FF_PROFILE_UNKNOWN, 0xffff, VE, "profile" },
82 #define PROFILE(name, value) name, NULL, 0, AV_OPT_TYPE_CONST, { .i64 = value }, 0, 0, VE, "profile"
83 { PROFILE("constrained_baseline", FF_PROFILE_H264_CONSTRAINED_BASELINE) },
84 { PROFILE("main", FF_PROFILE_H264_MAIN) },
85 { PROFILE("high", FF_PROFILE_H264_HIGH) },
87 { "max_nal_size", "set maximum NAL size in bytes", OFFSET(max_nal_size), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, VE },
88 { "allow_skip_frames", "allow skipping frames to hit the target bitrate", OFFSET(skip_frames), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
89 #if FF_API_OPENH264_CABAC
90 { "cabac", "Enable cabac(deprecated, use coder)", OFFSET(cabac), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE|DEPRECATED },
92 { "coder", "Coder type", OFFSET(coder), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 1, VE, "coder" },
93 { "default", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = -1 }, INT_MIN, INT_MAX, VE, "coder" },
94 { "cavlc", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = 0 }, INT_MIN, INT_MAX, VE, "coder" },
95 { "cabac", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = 1 }, INT_MIN, INT_MAX, VE, "coder" },
96 { "vlc", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = 0 }, INT_MIN, INT_MAX, VE, "coder" },
97 { "ac", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = 1 }, INT_MIN, INT_MAX, VE, "coder" },
99 { "rc_mode", "Select rate control mode", OFFSET(rc_mode), AV_OPT_TYPE_INT, { .i64 = RC_QUALITY_MODE }, RC_OFF_MODE, RC_TIMESTAMP_MODE, VE, "rc_mode" },
100 { "off", "bit rate control off", 0, AV_OPT_TYPE_CONST, { .i64 = RC_OFF_MODE }, 0, 0, VE, "rc_mode" },
101 { "quality", "quality mode", 0, AV_OPT_TYPE_CONST, { .i64 = RC_QUALITY_MODE }, 0, 0, VE, "rc_mode" },
102 { "bitrate", "bitrate mode", 0, AV_OPT_TYPE_CONST, { .i64 = RC_BITRATE_MODE }, 0, 0, VE, "rc_mode" },
103 { "buffer", "using buffer status to adjust the video quality (no bitrate control)", 0, AV_OPT_TYPE_CONST, { .i64 = RC_BUFFERBASED_MODE }, 0, 0, VE, "rc_mode" },
104 #if OPENH264_VER_AT_LEAST(1, 4)
105 { "timestamp", "bit rate control based on timestamp", 0, AV_OPT_TYPE_CONST, { .i64 = RC_TIMESTAMP_MODE }, 0, 0, VE, "rc_mode" },
111 static const AVClass class = {
112 .class_name = "libopenh264enc",
113 .item_name = av_default_item_name,
115 .version = LIBAVUTIL_VERSION_INT,
118 static av_cold int svc_encode_close(AVCodecContext *avctx)
120 SVCContext *s = avctx->priv_data;
123 WelsDestroySVCEncoder(s->encoder);
125 av_log(avctx, AV_LOG_WARNING, "%d frames skipped\n", s->skipped);
129 static av_cold int svc_encode_init(AVCodecContext *avctx)
131 SVCContext *s = avctx->priv_data;
132 SEncParamExt param = { 0 };
135 WelsTraceCallback callback_function;
136 AVCPBProperties *props;
138 if ((err = ff_libopenh264_check_version(avctx)) < 0)
141 if (WelsCreateSVCEncoder(&s->encoder)) {
142 av_log(avctx, AV_LOG_ERROR, "Unable to create encoder\n");
143 return AVERROR_UNKNOWN;
146 // Pass all libopenh264 messages to our callback, to allow ourselves to filter them.
147 log_level = WELS_LOG_DETAIL;
148 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_LEVEL, &log_level);
150 // Set the logging callback function to one that uses av_log() (see implementation above).
151 callback_function = (WelsTraceCallback) ff_libopenh264_trace_callback;
152 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_CALLBACK, &callback_function);
154 // Set the AVCodecContext as the libopenh264 callback context so that it can be passed to av_log().
155 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_CALLBACK_CONTEXT, &avctx);
157 (*s->encoder)->GetDefaultParams(s->encoder, ¶m);
159 param.fMaxFrameRate = 1/av_q2d(avctx->time_base);
160 param.iPicWidth = avctx->width;
161 param.iPicHeight = avctx->height;
162 param.iTargetBitrate = avctx->bit_rate > 0 ? avctx->bit_rate : TARGET_BITRATE_DEFAULT;
163 param.iMaxBitrate = FFMAX(avctx->rc_max_rate, avctx->bit_rate);
164 param.iRCMode = s->rc_mode;
165 if (avctx->qmax >= 0)
166 param.iMaxQp = av_clip(avctx->qmax, 1, 51);
167 if (avctx->qmin >= 0)
168 param.iMinQp = av_clip(avctx->qmin, 1, param.iMaxQp);
169 param.iTemporalLayerNum = 1;
170 param.iSpatialLayerNum = 1;
171 param.bEnableDenoise = 0;
172 param.bEnableBackgroundDetection = 1;
173 param.bEnableAdaptiveQuant = 1;
174 param.bEnableFrameSkip = s->skip_frames;
175 param.bEnableLongTermReference = 0;
176 param.iLtrMarkPeriod = 30;
177 if (avctx->gop_size >= 0)
178 param.uiIntraPeriod = avctx->gop_size;
179 #if OPENH264_VER_AT_LEAST(1, 4)
180 param.eSpsPpsIdStrategy = CONSTANT_ID;
182 param.bEnableSpsPpsIdAddition = 0;
184 param.bPrefixNalAddingCtrl = 0;
185 param.iLoopFilterDisableIdc = !s->loopfilter;
186 param.iEntropyCodingModeFlag = 0;
187 param.iMultipleThreadIdc = avctx->thread_count;
189 /* Allow specifying the libopenh264 profile through AVCodecContext. */
190 if (FF_PROFILE_UNKNOWN == s->profile &&
191 FF_PROFILE_UNKNOWN != avctx->profile)
192 switch (avctx->profile) {
193 case FF_PROFILE_H264_HIGH:
194 case FF_PROFILE_H264_MAIN:
195 case FF_PROFILE_H264_CONSTRAINED_BASELINE:
196 s->profile = avctx->profile;
199 av_log(avctx, AV_LOG_WARNING,
200 "Unsupported avctx->profile: %d.\n", avctx->profile);
204 #if FF_API_CODER_TYPE && FF_API_OPENH264_CABAC
205 FF_DISABLE_DEPRECATION_WARNINGS
206 if (s->coder < 0 && avctx->coder_type == FF_CODER_TYPE_AC)
211 FF_ENABLE_DEPRECATION_WARNINGS
214 if (s->profile == FF_PROFILE_UNKNOWN && s->coder >= 0)
215 s->profile = s->coder == 0 ? FF_PROFILE_H264_CONSTRAINED_BASELINE :
216 #if OPENH264_VER_AT_LEAST(1, 8)
217 FF_PROFILE_H264_HIGH;
219 FF_PROFILE_H264_MAIN;
222 switch (s->profile) {
223 #if OPENH264_VER_AT_LEAST(1, 8)
224 case FF_PROFILE_H264_HIGH:
225 param.iEntropyCodingModeFlag = 1;
226 av_log(avctx, AV_LOG_VERBOSE, "Using CABAC, "
227 "select EProfileIdc PRO_HIGH in libopenh264.\n");
230 case FF_PROFILE_H264_MAIN:
231 param.iEntropyCodingModeFlag = 1;
232 av_log(avctx, AV_LOG_VERBOSE, "Using CABAC, "
233 "select EProfileIdc PRO_MAIN in libopenh264.\n");
236 case FF_PROFILE_H264_CONSTRAINED_BASELINE:
237 case FF_PROFILE_UNKNOWN:
238 param.iEntropyCodingModeFlag = 0;
239 av_log(avctx, AV_LOG_VERBOSE, "Using CAVLC, "
240 "select EProfileIdc PRO_BASELINE in libopenh264.\n");
243 param.iEntropyCodingModeFlag = 0;
244 av_log(avctx, AV_LOG_WARNING, "Unsupported profile, "
245 "select EProfileIdc PRO_BASELINE in libopenh264.\n");
249 param.sSpatialLayers[0].iVideoWidth = param.iPicWidth;
250 param.sSpatialLayers[0].iVideoHeight = param.iPicHeight;
251 param.sSpatialLayers[0].fFrameRate = param.fMaxFrameRate;
252 param.sSpatialLayers[0].iSpatialBitrate = param.iTargetBitrate;
253 param.sSpatialLayers[0].iMaxSpatialBitrate = param.iMaxBitrate;
255 #if OPENH264_VER_AT_LEAST(1, 7)
256 if (avctx->sample_aspect_ratio.num && avctx->sample_aspect_ratio.den) {
258 static const AVRational sar_idc[] = {
259 { 0, 0 }, // Unspecified (never written here).
260 { 1, 1 }, { 12, 11 }, { 10, 11 }, { 16, 11 },
261 { 40, 33 }, { 24, 11 }, { 20, 11 }, { 32, 11 },
262 { 80, 33 }, { 18, 11 }, { 15, 11 }, { 64, 33 },
263 { 160, 99 }, // Last 3 are unknown to openh264: { 4, 3 }, { 3, 2 }, { 2, 1 },
265 static const ESampleAspectRatio asp_idc[] = {
267 ASP_1x1, ASP_12x11, ASP_10x11, ASP_16x11,
268 ASP_40x33, ASP_24x11, ASP_20x11, ASP_32x11,
269 ASP_80x33, ASP_18x11, ASP_15x11, ASP_64x33,
274 av_reduce(&num, &den, avctx->sample_aspect_ratio.num,
275 avctx->sample_aspect_ratio.den, 65535);
277 for (i = 1; i < FF_ARRAY_ELEMS(sar_idc); i++) {
278 if (num == sar_idc[i].num &&
279 den == sar_idc[i].den)
282 if (i == FF_ARRAY_ELEMS(sar_idc)) {
283 param.sSpatialLayers[0].eAspectRatio = ASP_EXT_SAR;
284 param.sSpatialLayers[0].sAspectRatioExtWidth = num;
285 param.sSpatialLayers[0].sAspectRatioExtHeight = den;
287 param.sSpatialLayers[0].eAspectRatio = asp_idc[i];
289 param.sSpatialLayers[0].bAspectRatioPresent = true;
291 param.sSpatialLayers[0].bAspectRatioPresent = false;
295 if ((avctx->slices > 1) && (s->max_nal_size)) {
296 av_log(avctx, AV_LOG_ERROR,
297 "Invalid combination -slices %d and -max_nal_size %d.\n",
298 avctx->slices, s->max_nal_size);
299 return AVERROR(EINVAL);
302 if (avctx->slices > 1)
303 s->slice_mode = SM_FIXEDSLCNUM_SLICE;
306 s->slice_mode = SM_SIZELIMITED_SLICE;
308 #if OPENH264_VER_AT_LEAST(1, 6)
309 param.sSpatialLayers[0].sSliceArgument.uiSliceMode = s->slice_mode;
310 param.sSpatialLayers[0].sSliceArgument.uiSliceNum = avctx->slices;
312 param.sSpatialLayers[0].sSliceCfg.uiSliceMode = s->slice_mode;
313 param.sSpatialLayers[0].sSliceCfg.sSliceArgument.uiSliceNum = avctx->slices;
315 if (avctx->slices == 0 && s->slice_mode == SM_FIXEDSLCNUM_SLICE)
316 av_log(avctx, AV_LOG_WARNING, "Slice count will be set automatically\n");
318 if (s->slice_mode == SM_SIZELIMITED_SLICE) {
319 if (s->max_nal_size) {
320 param.uiMaxNalSize = s->max_nal_size;
321 #if OPENH264_VER_AT_LEAST(1, 6)
322 param.sSpatialLayers[0].sSliceArgument.uiSliceSizeConstraint = s->max_nal_size;
324 param.sSpatialLayers[0].sSliceCfg.sSliceArgument.uiSliceSizeConstraint = s->max_nal_size;
327 av_log(avctx, AV_LOG_ERROR, "Invalid -max_nal_size, "
328 "specify a valid max_nal_size to use -slice_mode dyn\n");
329 return AVERROR(EINVAL);
333 if ((*s->encoder)->InitializeExt(s->encoder, ¶m) != cmResultSuccess) {
334 av_log(avctx, AV_LOG_ERROR, "Initialize failed\n");
335 return AVERROR_UNKNOWN;
338 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
339 SFrameBSInfo fbi = { 0 };
341 (*s->encoder)->EncodeParameterSets(s->encoder, &fbi);
342 for (i = 0; i < fbi.sLayerInfo[0].iNalCount; i++)
343 size += fbi.sLayerInfo[0].pNalLengthInByte[i];
344 avctx->extradata = av_mallocz(size + AV_INPUT_BUFFER_PADDING_SIZE);
345 if (!avctx->extradata)
346 return AVERROR(ENOMEM);
347 avctx->extradata_size = size;
348 memcpy(avctx->extradata, fbi.sLayerInfo[0].pBsBuf, size);
351 props = ff_add_cpb_side_data(avctx);
353 return AVERROR(ENOMEM);
354 props->max_bitrate = param.iMaxBitrate;
355 props->avg_bitrate = param.iTargetBitrate;
360 static int svc_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
361 const AVFrame *frame, int *got_packet)
363 SVCContext *s = avctx->priv_data;
364 SFrameBSInfo fbi = { 0 };
367 SSourcePicture sp = { 0 };
368 int size = 0, layer, first_layer = 0;
369 int layer_size[MAX_LAYER_NUM_OF_FRAME] = { 0 };
371 sp.iColorFormat = videoFormatI420;
372 for (i = 0; i < 3; i++) {
373 sp.iStride[i] = frame->linesize[i];
374 sp.pData[i] = frame->data[i];
376 sp.iPicWidth = avctx->width;
377 sp.iPicHeight = avctx->height;
379 if (frame->pict_type == AV_PICTURE_TYPE_I) {
380 (*s->encoder)->ForceIntraFrame(s->encoder, true);
383 encoded = (*s->encoder)->EncodeFrame(s->encoder, &sp, &fbi);
384 if (encoded != cmResultSuccess) {
385 av_log(avctx, AV_LOG_ERROR, "EncodeFrame failed\n");
386 return AVERROR_UNKNOWN;
388 if (fbi.eFrameType == videoFrameTypeSkip) {
390 av_log(avctx, AV_LOG_DEBUG, "frame skipped\n");
394 // Normal frames are returned with one single layer, while IDR
395 // frames have two layers, where the first layer contains the SPS/PPS.
396 // If using global headers, don't include the SPS/PPS in the returned
397 // packet - thus, only return one layer.
398 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER)
399 first_layer = fbi.iLayerNum - 1;
401 for (layer = first_layer; layer < fbi.iLayerNum; layer++) {
402 for (i = 0; i < fbi.sLayerInfo[layer].iNalCount; i++)
403 layer_size[layer] += fbi.sLayerInfo[layer].pNalLengthInByte[i];
404 size += layer_size[layer];
406 av_log(avctx, AV_LOG_DEBUG, "%d slices\n", fbi.sLayerInfo[fbi.iLayerNum - 1].iNalCount);
408 if ((ret = ff_alloc_packet2(avctx, avpkt, size, size))) {
409 av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
413 for (layer = first_layer; layer < fbi.iLayerNum; layer++) {
414 memcpy(avpkt->data + size, fbi.sLayerInfo[layer].pBsBuf, layer_size[layer]);
415 size += layer_size[layer];
417 avpkt->pts = frame->pts;
418 if (fbi.eFrameType == videoFrameTypeIDR)
419 avpkt->flags |= AV_PKT_FLAG_KEY;
424 static const AVCodecDefault svc_enc_defaults[] = {
432 AVCodec ff_libopenh264_encoder = {
433 .name = "libopenh264",
434 .long_name = NULL_IF_CONFIG_SMALL("OpenH264 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
435 .type = AVMEDIA_TYPE_VIDEO,
436 .id = AV_CODEC_ID_H264,
437 .priv_data_size = sizeof(SVCContext),
438 .init = svc_encode_init,
439 .encode2 = svc_encode_frame,
440 .close = svc_encode_close,
441 .capabilities = AV_CODEC_CAP_AUTO_THREADS,
442 .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE | FF_CODEC_CAP_INIT_CLEANUP,
443 .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P,
445 .defaults = svc_enc_defaults,
446 .priv_class = &class,
447 .wrapper_name = "libopenh264",