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;
57 #define OFFSET(x) offsetof(SVCContext, x)
58 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
59 #define DEPRECATED AV_OPT_FLAG_DEPRECATED
60 static const AVOption options[] = {
61 #if FF_API_OPENH264_SLICE_MODE
62 #if OPENH264_VER_AT_LEAST(1, 6)
63 { "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" },
65 { "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" },
67 { "fixed", "a fixed number of slices", 0, AV_OPT_TYPE_CONST, { .i64 = SM_FIXEDSLCNUM_SLICE }, 0, 0, VE, "slice_mode" },
68 #if OPENH264_VER_AT_LEAST(1, 6)
69 { "dyn", "Size limited (compatibility name)", 0, AV_OPT_TYPE_CONST, { .i64 = SM_SIZELIMITED_SLICE }, 0, 0, VE, "slice_mode" },
70 { "sizelimited", "Size limited", 0, AV_OPT_TYPE_CONST, { .i64 = SM_SIZELIMITED_SLICE }, 0, 0, VE, "slice_mode" },
72 { "rowmb", "one slice per row of macroblocks", 0, AV_OPT_TYPE_CONST, { .i64 = SM_ROWMB_SLICE }, 0, 0, VE, "slice_mode" },
73 { "auto", "automatic number of slices according to number of threads", 0, AV_OPT_TYPE_CONST, { .i64 = SM_AUTO_SLICE }, 0, 0, VE, "slice_mode" },
74 { "dyn", "Dynamic slicing", 0, AV_OPT_TYPE_CONST, { .i64 = SM_DYN_SLICE }, 0, 0, VE, "slice_mode" },
77 { "loopfilter", "enable loop filter", OFFSET(loopfilter), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, VE },
78 { "profile", "set profile restrictions", OFFSET(profile), AV_OPT_TYPE_INT, { .i64 = FF_PROFILE_UNKNOWN }, FF_PROFILE_UNKNOWN, 0xffff, VE, "profile" },
79 #define PROFILE(name, value) name, NULL, 0, AV_OPT_TYPE_CONST, { .i64 = value }, 0, 0, VE, "profile"
80 { PROFILE("constrained_baseline", FF_PROFILE_H264_CONSTRAINED_BASELINE) },
81 { PROFILE("main", FF_PROFILE_H264_MAIN) },
82 { PROFILE("high", FF_PROFILE_H264_HIGH) },
84 { "max_nal_size", "set maximum NAL size in bytes", OFFSET(max_nal_size), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, VE },
85 { "allow_skip_frames", "allow skipping frames to hit the target bitrate", OFFSET(skip_frames), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
86 { "cabac", "Enable cabac", OFFSET(cabac), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE },
88 { "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" },
89 { "off", "bit rate control off", 0, AV_OPT_TYPE_CONST, { .i64 = RC_OFF_MODE }, 0, 0, VE, "rc_mode" },
90 { "quality", "quality mode", 0, AV_OPT_TYPE_CONST, { .i64 = RC_QUALITY_MODE }, 0, 0, VE, "rc_mode" },
91 { "bitrate", "bitrate mode", 0, AV_OPT_TYPE_CONST, { .i64 = RC_BITRATE_MODE }, 0, 0, VE, "rc_mode" },
92 { "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" },
93 #if OPENH264_VER_AT_LEAST(1, 4)
94 { "timestamp", "bit rate control based on timestamp", 0, AV_OPT_TYPE_CONST, { .i64 = RC_TIMESTAMP_MODE }, 0, 0, VE, "rc_mode" },
100 static const AVClass class = {
101 .class_name = "libopenh264enc",
102 .item_name = av_default_item_name,
104 .version = LIBAVUTIL_VERSION_INT,
107 static av_cold int svc_encode_close(AVCodecContext *avctx)
109 SVCContext *s = avctx->priv_data;
112 WelsDestroySVCEncoder(s->encoder);
114 av_log(avctx, AV_LOG_WARNING, "%d frames skipped\n", s->skipped);
118 static av_cold int svc_encode_init(AVCodecContext *avctx)
120 SVCContext *s = avctx->priv_data;
121 SEncParamExt param = { 0 };
124 WelsTraceCallback callback_function;
125 AVCPBProperties *props;
127 if ((err = ff_libopenh264_check_version(avctx)) < 0)
130 if (WelsCreateSVCEncoder(&s->encoder)) {
131 av_log(avctx, AV_LOG_ERROR, "Unable to create encoder\n");
132 return AVERROR_UNKNOWN;
135 // Pass all libopenh264 messages to our callback, to allow ourselves to filter them.
136 log_level = WELS_LOG_DETAIL;
137 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_LEVEL, &log_level);
139 // Set the logging callback function to one that uses av_log() (see implementation above).
140 callback_function = (WelsTraceCallback) ff_libopenh264_trace_callback;
141 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_CALLBACK, &callback_function);
143 // Set the AVCodecContext as the libopenh264 callback context so that it can be passed to av_log().
144 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_CALLBACK_CONTEXT, &avctx);
146 (*s->encoder)->GetDefaultParams(s->encoder, ¶m);
148 #if FF_API_CODER_TYPE
149 FF_DISABLE_DEPRECATION_WARNINGS
151 s->cabac = avctx->coder_type == FF_CODER_TYPE_AC;
152 FF_ENABLE_DEPRECATION_WARNINGS
155 param.fMaxFrameRate = 1/av_q2d(avctx->time_base);
156 param.iPicWidth = avctx->width;
157 param.iPicHeight = avctx->height;
158 param.iTargetBitrate = avctx->bit_rate > 0 ? avctx->bit_rate : TARGET_BITRATE_DEFAULT;
159 param.iMaxBitrate = FFMAX(avctx->rc_max_rate, avctx->bit_rate);
160 param.iRCMode = s->rc_mode;
161 if (avctx->qmax >= 0)
162 param.iMaxQp = av_clip(avctx->qmax, 1, 51);
163 if (avctx->qmin >= 0)
164 param.iMinQp = av_clip(avctx->qmin, 1, param.iMaxQp);
165 param.iTemporalLayerNum = 1;
166 param.iSpatialLayerNum = 1;
167 param.bEnableDenoise = 0;
168 param.bEnableBackgroundDetection = 1;
169 param.bEnableAdaptiveQuant = 1;
170 param.bEnableFrameSkip = s->skip_frames;
171 param.bEnableLongTermReference = 0;
172 param.iLtrMarkPeriod = 30;
173 if (avctx->gop_size >= 0)
174 param.uiIntraPeriod = avctx->gop_size;
175 #if OPENH264_VER_AT_LEAST(1, 4)
176 param.eSpsPpsIdStrategy = CONSTANT_ID;
178 param.bEnableSpsPpsIdAddition = 0;
180 param.bPrefixNalAddingCtrl = 0;
181 param.iLoopFilterDisableIdc = !s->loopfilter;
182 param.iEntropyCodingModeFlag = 0;
183 param.iMultipleThreadIdc = avctx->thread_count;
185 if (s->profile == FF_PROFILE_UNKNOWN)
186 s->profile = !s->cabac ? FF_PROFILE_H264_CONSTRAINED_BASELINE :
187 #if OPENH264_VER_AT_LEAST(1, 8)
188 FF_PROFILE_H264_HIGH;
190 FF_PROFILE_H264_MAIN;
193 switch (s->profile) {
194 #if OPENH264_VER_AT_LEAST(1, 8)
195 case FF_PROFILE_H264_HIGH:
196 param.iEntropyCodingModeFlag = 1;
197 av_log(avctx, AV_LOG_VERBOSE, "Using CABAC, "
198 "select EProfileIdc PRO_HIGH in libopenh264.\n");
201 case FF_PROFILE_H264_MAIN:
202 param.iEntropyCodingModeFlag = 1;
203 av_log(avctx, AV_LOG_VERBOSE, "Using CABAC, "
204 "select EProfileIdc PRO_MAIN in libopenh264.\n");
207 case FF_PROFILE_H264_CONSTRAINED_BASELINE:
208 case FF_PROFILE_UNKNOWN:
209 param.iEntropyCodingModeFlag = 0;
210 av_log(avctx, AV_LOG_VERBOSE, "Using CAVLC, "
211 "select EProfileIdc PRO_BASELINE in libopenh264.\n");
214 param.iEntropyCodingModeFlag = 0;
215 av_log(avctx, AV_LOG_WARNING, "Unsupported profile, "
216 "select EProfileIdc PRO_BASELINE in libopenh264.\n");
220 param.sSpatialLayers[0].iVideoWidth = param.iPicWidth;
221 param.sSpatialLayers[0].iVideoHeight = param.iPicHeight;
222 param.sSpatialLayers[0].fFrameRate = param.fMaxFrameRate;
223 param.sSpatialLayers[0].iSpatialBitrate = param.iTargetBitrate;
224 param.sSpatialLayers[0].iMaxSpatialBitrate = param.iMaxBitrate;
226 #if OPENH264_VER_AT_LEAST(1, 7)
227 if (avctx->sample_aspect_ratio.num && avctx->sample_aspect_ratio.den) {
229 static const AVRational sar_idc[] = {
230 { 0, 0 }, // Unspecified (never written here).
231 { 1, 1 }, { 12, 11 }, { 10, 11 }, { 16, 11 },
232 { 40, 33 }, { 24, 11 }, { 20, 11 }, { 32, 11 },
233 { 80, 33 }, { 18, 11 }, { 15, 11 }, { 64, 33 },
234 { 160, 99 }, // Last 3 are unknown to openh264: { 4, 3 }, { 3, 2 }, { 2, 1 },
236 static const ESampleAspectRatio asp_idc[] = {
238 ASP_1x1, ASP_12x11, ASP_10x11, ASP_16x11,
239 ASP_40x33, ASP_24x11, ASP_20x11, ASP_32x11,
240 ASP_80x33, ASP_18x11, ASP_15x11, ASP_64x33,
245 av_reduce(&num, &den, avctx->sample_aspect_ratio.num,
246 avctx->sample_aspect_ratio.den, 65535);
248 for (i = 1; i < FF_ARRAY_ELEMS(sar_idc); i++) {
249 if (num == sar_idc[i].num &&
250 den == sar_idc[i].den)
253 if (i == FF_ARRAY_ELEMS(sar_idc)) {
254 param.sSpatialLayers[0].eAspectRatio = ASP_EXT_SAR;
255 param.sSpatialLayers[0].sAspectRatioExtWidth = num;
256 param.sSpatialLayers[0].sAspectRatioExtHeight = den;
258 param.sSpatialLayers[0].eAspectRatio = asp_idc[i];
260 param.sSpatialLayers[0].bAspectRatioPresent = true;
262 param.sSpatialLayers[0].bAspectRatioPresent = false;
266 if ((avctx->slices > 1) && (s->max_nal_size)) {
267 av_log(avctx, AV_LOG_ERROR,
268 "Invalid combination -slices %d and -max_nal_size %d.\n",
269 avctx->slices, s->max_nal_size);
270 return AVERROR(EINVAL);
273 if (avctx->slices > 1)
274 s->slice_mode = SM_FIXEDSLCNUM_SLICE;
277 s->slice_mode = SM_SIZELIMITED_SLICE;
279 #if OPENH264_VER_AT_LEAST(1, 6)
280 param.sSpatialLayers[0].sSliceArgument.uiSliceMode = s->slice_mode;
281 param.sSpatialLayers[0].sSliceArgument.uiSliceNum = avctx->slices;
283 param.sSpatialLayers[0].sSliceCfg.uiSliceMode = s->slice_mode;
284 param.sSpatialLayers[0].sSliceCfg.sSliceArgument.uiSliceNum = avctx->slices;
286 if (avctx->slices == 0 && s->slice_mode == SM_FIXEDSLCNUM_SLICE)
287 av_log(avctx, AV_LOG_WARNING, "Slice count will be set automatically\n");
289 if (s->slice_mode == SM_SIZELIMITED_SLICE) {
290 if (s->max_nal_size) {
291 param.uiMaxNalSize = s->max_nal_size;
292 #if OPENH264_VER_AT_LEAST(1, 6)
293 param.sSpatialLayers[0].sSliceArgument.uiSliceSizeConstraint = s->max_nal_size;
295 param.sSpatialLayers[0].sSliceCfg.sSliceArgument.uiSliceSizeConstraint = s->max_nal_size;
298 av_log(avctx, AV_LOG_ERROR, "Invalid -max_nal_size, "
299 "specify a valid max_nal_size to use -slice_mode dyn\n");
300 return AVERROR(EINVAL);
304 if ((*s->encoder)->InitializeExt(s->encoder, ¶m) != cmResultSuccess) {
305 av_log(avctx, AV_LOG_ERROR, "Initialize failed\n");
306 return AVERROR_UNKNOWN;
309 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
310 SFrameBSInfo fbi = { 0 };
312 (*s->encoder)->EncodeParameterSets(s->encoder, &fbi);
313 for (i = 0; i < fbi.sLayerInfo[0].iNalCount; i++)
314 size += fbi.sLayerInfo[0].pNalLengthInByte[i];
315 avctx->extradata = av_mallocz(size + AV_INPUT_BUFFER_PADDING_SIZE);
316 if (!avctx->extradata)
317 return AVERROR(ENOMEM);
318 avctx->extradata_size = size;
319 memcpy(avctx->extradata, fbi.sLayerInfo[0].pBsBuf, size);
322 props = ff_add_cpb_side_data(avctx);
324 return AVERROR(ENOMEM);
325 props->max_bitrate = param.iMaxBitrate;
326 props->avg_bitrate = param.iTargetBitrate;
331 static int svc_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
332 const AVFrame *frame, int *got_packet)
334 SVCContext *s = avctx->priv_data;
335 SFrameBSInfo fbi = { 0 };
338 SSourcePicture sp = { 0 };
339 int size = 0, layer, first_layer = 0;
340 int layer_size[MAX_LAYER_NUM_OF_FRAME] = { 0 };
342 sp.iColorFormat = videoFormatI420;
343 for (i = 0; i < 3; i++) {
344 sp.iStride[i] = frame->linesize[i];
345 sp.pData[i] = frame->data[i];
347 sp.iPicWidth = avctx->width;
348 sp.iPicHeight = avctx->height;
350 if (frame->pict_type == AV_PICTURE_TYPE_I) {
351 (*s->encoder)->ForceIntraFrame(s->encoder, true);
354 encoded = (*s->encoder)->EncodeFrame(s->encoder, &sp, &fbi);
355 if (encoded != cmResultSuccess) {
356 av_log(avctx, AV_LOG_ERROR, "EncodeFrame failed\n");
357 return AVERROR_UNKNOWN;
359 if (fbi.eFrameType == videoFrameTypeSkip) {
361 av_log(avctx, AV_LOG_DEBUG, "frame skipped\n");
365 // Normal frames are returned with one single layer, while IDR
366 // frames have two layers, where the first layer contains the SPS/PPS.
367 // If using global headers, don't include the SPS/PPS in the returned
368 // packet - thus, only return one layer.
369 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER)
370 first_layer = fbi.iLayerNum - 1;
372 for (layer = first_layer; layer < fbi.iLayerNum; layer++) {
373 for (i = 0; i < fbi.sLayerInfo[layer].iNalCount; i++)
374 layer_size[layer] += fbi.sLayerInfo[layer].pNalLengthInByte[i];
375 size += layer_size[layer];
377 av_log(avctx, AV_LOG_DEBUG, "%d slices\n", fbi.sLayerInfo[fbi.iLayerNum - 1].iNalCount);
379 if ((ret = ff_alloc_packet2(avctx, avpkt, size, size))) {
380 av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
384 for (layer = first_layer; layer < fbi.iLayerNum; layer++) {
385 memcpy(avpkt->data + size, fbi.sLayerInfo[layer].pBsBuf, layer_size[layer]);
386 size += layer_size[layer];
388 avpkt->pts = frame->pts;
389 if (fbi.eFrameType == videoFrameTypeIDR)
390 avpkt->flags |= AV_PKT_FLAG_KEY;
395 static const AVCodecDefault svc_enc_defaults[] = {
403 AVCodec ff_libopenh264_encoder = {
404 .name = "libopenh264",
405 .long_name = NULL_IF_CONFIG_SMALL("OpenH264 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
406 .type = AVMEDIA_TYPE_VIDEO,
407 .id = AV_CODEC_ID_H264,
408 .priv_data_size = sizeof(SVCContext),
409 .init = svc_encode_init,
410 .encode2 = svc_encode_frame,
411 .close = svc_encode_close,
412 .capabilities = AV_CODEC_CAP_AUTO_THREADS,
413 .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE | FF_CODEC_CAP_INIT_CLEANUP,
414 .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P,
416 .defaults = svc_enc_defaults,
417 .priv_class = &class,
418 .wrapper_name = "libopenh264",