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;
54 #define OFFSET(x) offsetof(SVCContext, x)
55 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
56 static const AVOption options[] = {
57 #if OPENH264_VER_AT_LEAST(1, 6)
58 { "slice_mode", "set slice mode", OFFSET(slice_mode), AV_OPT_TYPE_INT, { .i64 = SM_FIXEDSLCNUM_SLICE }, SM_SINGLE_SLICE, SM_RESERVED, VE, "slice_mode" },
60 { "slice_mode", "set slice mode", OFFSET(slice_mode), AV_OPT_TYPE_INT, { .i64 = SM_AUTO_SLICE }, SM_SINGLE_SLICE, SM_RESERVED, VE, "slice_mode" },
62 { "fixed", "a fixed number of slices", 0, AV_OPT_TYPE_CONST, { .i64 = SM_FIXEDSLCNUM_SLICE }, 0, 0, VE, "slice_mode" },
63 #if OPENH264_VER_AT_LEAST(1, 6)
64 { "dyn", "Size limited (compatibility name)", 0, AV_OPT_TYPE_CONST, { .i64 = SM_SIZELIMITED_SLICE }, 0, 0, VE, "slice_mode" },
65 { "sizelimited", "Size limited", 0, AV_OPT_TYPE_CONST, { .i64 = SM_SIZELIMITED_SLICE }, 0, 0, VE, "slice_mode" },
67 { "rowmb", "one slice per row of macroblocks", 0, AV_OPT_TYPE_CONST, { .i64 = SM_ROWMB_SLICE }, 0, 0, VE, "slice_mode" },
68 { "auto", "automatic number of slices according to number of threads", 0, AV_OPT_TYPE_CONST, { .i64 = SM_AUTO_SLICE }, 0, 0, VE, "slice_mode" },
69 { "dyn", "Dynamic slicing", 0, AV_OPT_TYPE_CONST, { .i64 = SM_DYN_SLICE }, 0, 0, VE, "slice_mode" },
71 { "loopfilter", "enable loop filter", OFFSET(loopfilter), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, VE },
72 { "profile", "set profile restrictions", OFFSET(profile), AV_OPT_TYPE_STRING, { .str = NULL }, 0, 0, VE },
73 { "max_nal_size", "set maximum NAL size in bytes", OFFSET(max_nal_size), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, VE },
74 { "allow_skip_frames", "allow skipping frames to hit the target bitrate", OFFSET(skip_frames), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
75 { "cabac", "Enable cabac", OFFSET(cabac), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE },
79 static const AVClass class = {
80 .class_name = "libopenh264enc",
81 .item_name = av_default_item_name,
83 .version = LIBAVUTIL_VERSION_INT,
86 static av_cold int svc_encode_close(AVCodecContext *avctx)
88 SVCContext *s = avctx->priv_data;
91 WelsDestroySVCEncoder(s->encoder);
93 av_log(avctx, AV_LOG_WARNING, "%d frames skipped\n", s->skipped);
97 static av_cold int svc_encode_init(AVCodecContext *avctx)
99 SVCContext *s = avctx->priv_data;
100 SEncParamExt param = { 0 };
103 WelsTraceCallback callback_function;
104 AVCPBProperties *props;
106 if ((err = ff_libopenh264_check_version(avctx)) < 0)
109 if (WelsCreateSVCEncoder(&s->encoder)) {
110 av_log(avctx, AV_LOG_ERROR, "Unable to create encoder\n");
111 return AVERROR_UNKNOWN;
114 // Pass all libopenh264 messages to our callback, to allow ourselves to filter them.
115 log_level = WELS_LOG_DETAIL;
116 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_LEVEL, &log_level);
118 // Set the logging callback function to one that uses av_log() (see implementation above).
119 callback_function = (WelsTraceCallback) ff_libopenh264_trace_callback;
120 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_CALLBACK, &callback_function);
122 // Set the AVCodecContext as the libopenh264 callback context so that it can be passed to av_log().
123 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_CALLBACK_CONTEXT, &avctx);
125 (*s->encoder)->GetDefaultParams(s->encoder, ¶m);
127 #if FF_API_CODER_TYPE
128 FF_DISABLE_DEPRECATION_WARNINGS
130 s->cabac = avctx->coder_type == FF_CODER_TYPE_AC;
131 FF_ENABLE_DEPRECATION_WARNINGS
134 param.fMaxFrameRate = 1/av_q2d(avctx->time_base);
135 param.iPicWidth = avctx->width;
136 param.iPicHeight = avctx->height;
137 param.iTargetBitrate = avctx->bit_rate > 0 ? avctx->bit_rate : TARGET_BITRATE_DEFAULT;
138 param.iMaxBitrate = FFMAX(avctx->rc_max_rate, avctx->bit_rate);
139 param.iRCMode = RC_QUALITY_MODE;
140 if (avctx->qmax >= 0)
141 param.iMaxQp = av_clip(avctx->qmax, 1, 51);
142 if (avctx->qmin >= 0)
143 param.iMinQp = av_clip(avctx->qmin, 1, param.iMaxQp);
144 param.iTemporalLayerNum = 1;
145 param.iSpatialLayerNum = 1;
146 param.bEnableDenoise = 0;
147 param.bEnableBackgroundDetection = 1;
148 param.bEnableAdaptiveQuant = 1;
149 param.bEnableFrameSkip = s->skip_frames;
150 param.bEnableLongTermReference = 0;
151 param.iLtrMarkPeriod = 30;
152 if (avctx->gop_size >= 0)
153 param.uiIntraPeriod = avctx->gop_size;
154 #if OPENH264_VER_AT_LEAST(1, 4)
155 param.eSpsPpsIdStrategy = CONSTANT_ID;
157 param.bEnableSpsPpsIdAddition = 0;
159 param.bPrefixNalAddingCtrl = 0;
160 param.iLoopFilterDisableIdc = !s->loopfilter;
161 param.iEntropyCodingModeFlag = 0;
162 param.iMultipleThreadIdc = avctx->thread_count;
163 if (s->profile && !strcmp(s->profile, "main"))
164 param.iEntropyCodingModeFlag = 1;
165 else if (!s->profile && s->cabac)
166 param.iEntropyCodingModeFlag = 1;
168 param.sSpatialLayers[0].iVideoWidth = param.iPicWidth;
169 param.sSpatialLayers[0].iVideoHeight = param.iPicHeight;
170 param.sSpatialLayers[0].fFrameRate = param.fMaxFrameRate;
171 param.sSpatialLayers[0].iSpatialBitrate = param.iTargetBitrate;
172 param.sSpatialLayers[0].iMaxSpatialBitrate = param.iMaxBitrate;
174 #if OPENH264_VER_AT_LEAST(1, 7)
175 if (avctx->sample_aspect_ratio.num && avctx->sample_aspect_ratio.den) {
177 static const AVRational sar_idc[] = {
178 { 0, 0 }, // Unspecified (never written here).
179 { 1, 1 }, { 12, 11 }, { 10, 11 }, { 16, 11 },
180 { 40, 33 }, { 24, 11 }, { 20, 11 }, { 32, 11 },
181 { 80, 33 }, { 18, 11 }, { 15, 11 }, { 64, 33 },
182 { 160, 99 }, // Last 3 are unknown to openh264: { 4, 3 }, { 3, 2 }, { 2, 1 },
184 static const ESampleAspectRatio asp_idc[] = {
186 ASP_1x1, ASP_12x11, ASP_10x11, ASP_16x11,
187 ASP_40x33, ASP_24x11, ASP_20x11, ASP_32x11,
188 ASP_80x33, ASP_18x11, ASP_15x11, ASP_64x33,
193 av_reduce(&num, &den, avctx->sample_aspect_ratio.num,
194 avctx->sample_aspect_ratio.den, 65535);
196 for (i = 1; i < FF_ARRAY_ELEMS(sar_idc); i++) {
197 if (num == sar_idc[i].num &&
198 den == sar_idc[i].den)
201 if (i == FF_ARRAY_ELEMS(sar_idc)) {
202 param.sSpatialLayers[0].eAspectRatio = ASP_EXT_SAR;
203 param.sSpatialLayers[0].sAspectRatioExtWidth = num;
204 param.sSpatialLayers[0].sAspectRatioExtHeight = den;
206 param.sSpatialLayers[0].eAspectRatio = asp_idc[i];
208 param.sSpatialLayers[0].bAspectRatioPresent = true;
210 param.sSpatialLayers[0].bAspectRatioPresent = false;
214 if ((avctx->slices > 1) && (s->max_nal_size)) {
215 av_log(avctx, AV_LOG_ERROR,
216 "Invalid combination -slices %d and -max_nal_size %d.\n",
217 avctx->slices, s->max_nal_size);
218 return AVERROR(EINVAL);
221 if (avctx->slices > 1)
222 s->slice_mode = SM_FIXEDSLCNUM_SLICE;
225 s->slice_mode = SM_SIZELIMITED_SLICE;
227 #if OPENH264_VER_AT_LEAST(1, 6)
228 param.sSpatialLayers[0].sSliceArgument.uiSliceMode = s->slice_mode;
229 param.sSpatialLayers[0].sSliceArgument.uiSliceNum = avctx->slices;
231 param.sSpatialLayers[0].sSliceCfg.uiSliceMode = s->slice_mode;
232 param.sSpatialLayers[0].sSliceCfg.sSliceArgument.uiSliceNum = avctx->slices;
235 if (s->slice_mode == SM_SIZELIMITED_SLICE) {
236 if (s->max_nal_size) {
237 param.uiMaxNalSize = s->max_nal_size;
238 #if OPENH264_VER_AT_LEAST(1, 6)
239 param.sSpatialLayers[0].sSliceArgument.uiSliceSizeConstraint = s->max_nal_size;
241 param.sSpatialLayers[0].sSliceCfg.sSliceArgument.uiSliceSizeConstraint = s->max_nal_size;
244 av_log(avctx, AV_LOG_ERROR, "Invalid -max_nal_size, "
245 "specify a valid max_nal_size to use -slice_mode dyn\n");
246 return AVERROR(EINVAL);
250 if ((*s->encoder)->InitializeExt(s->encoder, ¶m) != cmResultSuccess) {
251 av_log(avctx, AV_LOG_ERROR, "Initialize failed\n");
252 return AVERROR_UNKNOWN;
255 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
256 SFrameBSInfo fbi = { 0 };
258 (*s->encoder)->EncodeParameterSets(s->encoder, &fbi);
259 for (i = 0; i < fbi.sLayerInfo[0].iNalCount; i++)
260 size += fbi.sLayerInfo[0].pNalLengthInByte[i];
261 avctx->extradata = av_mallocz(size + AV_INPUT_BUFFER_PADDING_SIZE);
262 if (!avctx->extradata)
263 return AVERROR(ENOMEM);
264 avctx->extradata_size = size;
265 memcpy(avctx->extradata, fbi.sLayerInfo[0].pBsBuf, size);
268 props = ff_add_cpb_side_data(avctx);
270 return AVERROR(ENOMEM);
271 props->max_bitrate = param.iMaxBitrate;
272 props->avg_bitrate = param.iTargetBitrate;
277 static int svc_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
278 const AVFrame *frame, int *got_packet)
280 SVCContext *s = avctx->priv_data;
281 SFrameBSInfo fbi = { 0 };
284 SSourcePicture sp = { 0 };
285 int size = 0, layer, first_layer = 0;
286 int layer_size[MAX_LAYER_NUM_OF_FRAME] = { 0 };
288 sp.iColorFormat = videoFormatI420;
289 for (i = 0; i < 3; i++) {
290 sp.iStride[i] = frame->linesize[i];
291 sp.pData[i] = frame->data[i];
293 sp.iPicWidth = avctx->width;
294 sp.iPicHeight = avctx->height;
296 if (frame->pict_type == AV_PICTURE_TYPE_I) {
297 (*s->encoder)->ForceIntraFrame(s->encoder, true);
300 encoded = (*s->encoder)->EncodeFrame(s->encoder, &sp, &fbi);
301 if (encoded != cmResultSuccess) {
302 av_log(avctx, AV_LOG_ERROR, "EncodeFrame failed\n");
303 return AVERROR_UNKNOWN;
305 if (fbi.eFrameType == videoFrameTypeSkip) {
307 av_log(avctx, AV_LOG_DEBUG, "frame skipped\n");
311 // Normal frames are returned with one single layer, while IDR
312 // frames have two layers, where the first layer contains the SPS/PPS.
313 // If using global headers, don't include the SPS/PPS in the returned
314 // packet - thus, only return one layer.
315 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER)
316 first_layer = fbi.iLayerNum - 1;
318 for (layer = first_layer; layer < fbi.iLayerNum; layer++) {
319 for (i = 0; i < fbi.sLayerInfo[layer].iNalCount; i++)
320 layer_size[layer] += fbi.sLayerInfo[layer].pNalLengthInByte[i];
321 size += layer_size[layer];
323 av_log(avctx, AV_LOG_DEBUG, "%d slices\n", fbi.sLayerInfo[fbi.iLayerNum - 1].iNalCount);
325 if ((ret = ff_alloc_packet2(avctx, avpkt, size, size))) {
326 av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
330 for (layer = first_layer; layer < fbi.iLayerNum; layer++) {
331 memcpy(avpkt->data + size, fbi.sLayerInfo[layer].pBsBuf, layer_size[layer]);
332 size += layer_size[layer];
334 avpkt->pts = frame->pts;
335 if (fbi.eFrameType == videoFrameTypeIDR)
336 avpkt->flags |= AV_PKT_FLAG_KEY;
341 static const AVCodecDefault svc_enc_defaults[] = {
349 AVCodec ff_libopenh264_encoder = {
350 .name = "libopenh264",
351 .long_name = NULL_IF_CONFIG_SMALL("OpenH264 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
352 .type = AVMEDIA_TYPE_VIDEO,
353 .id = AV_CODEC_ID_H264,
354 .priv_data_size = sizeof(SVCContext),
355 .init = svc_encode_init,
356 .encode2 = svc_encode_frame,
357 .close = svc_encode_close,
358 .capabilities = AV_CODEC_CAP_AUTO_THREADS,
359 .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE | FF_CODEC_CAP_INIT_CLEANUP,
360 .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P,
362 .defaults = svc_enc_defaults,
363 .priv_class = &class,
364 .wrapper_name = "libopenh264",