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 static const AVOption options[] = {
60 #if OPENH264_VER_AT_LEAST(1, 6)
61 { "slice_mode", "set slice mode", OFFSET(slice_mode), AV_OPT_TYPE_INT, { .i64 = SM_FIXEDSLCNUM_SLICE }, SM_SINGLE_SLICE, SM_RESERVED, VE, "slice_mode" },
63 { "slice_mode", "set slice mode", OFFSET(slice_mode), AV_OPT_TYPE_INT, { .i64 = SM_AUTO_SLICE }, SM_SINGLE_SLICE, SM_RESERVED, VE, "slice_mode" },
65 { "fixed", "a fixed number of slices", 0, AV_OPT_TYPE_CONST, { .i64 = SM_FIXEDSLCNUM_SLICE }, 0, 0, VE, "slice_mode" },
66 #if OPENH264_VER_AT_LEAST(1, 6)
67 { "dyn", "Size limited (compatibility name)", 0, AV_OPT_TYPE_CONST, { .i64 = SM_SIZELIMITED_SLICE }, 0, 0, VE, "slice_mode" },
68 { "sizelimited", "Size limited", 0, AV_OPT_TYPE_CONST, { .i64 = SM_SIZELIMITED_SLICE }, 0, 0, VE, "slice_mode" },
70 { "rowmb", "one slice per row of macroblocks", 0, AV_OPT_TYPE_CONST, { .i64 = SM_ROWMB_SLICE }, 0, 0, VE, "slice_mode" },
71 { "auto", "automatic number of slices according to number of threads", 0, AV_OPT_TYPE_CONST, { .i64 = SM_AUTO_SLICE }, 0, 0, VE, "slice_mode" },
72 { "dyn", "Dynamic slicing", 0, AV_OPT_TYPE_CONST, { .i64 = SM_DYN_SLICE }, 0, 0, VE, "slice_mode" },
74 { "loopfilter", "enable loop filter", OFFSET(loopfilter), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, VE },
75 { "profile", "set profile restrictions", OFFSET(profile), AV_OPT_TYPE_STRING, { .str = NULL }, 0, 0, VE },
76 { "max_nal_size", "set maximum NAL size in bytes", OFFSET(max_nal_size), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, VE },
77 { "allow_skip_frames", "allow skipping frames to hit the target bitrate", OFFSET(skip_frames), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
78 { "cabac", "Enable cabac", OFFSET(cabac), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE },
80 { "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" },
81 { "off", "bit rate control off", 0, AV_OPT_TYPE_CONST, { .i64 = RC_OFF_MODE }, 0, 0, VE, "rc_mode" },
82 { "quality", "quality mode", 0, AV_OPT_TYPE_CONST, { .i64 = RC_QUALITY_MODE }, 0, 0, VE, "rc_mode" },
83 { "bitrate", "bitrate mode", 0, AV_OPT_TYPE_CONST, { .i64 = RC_BITRATE_MODE }, 0, 0, VE, "rc_mode" },
84 { "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" },
85 #if OPENH264_VER_AT_LEAST(1, 4)
86 { "timestamp", "bit rate control based on timestamp", 0, AV_OPT_TYPE_CONST, { .i64 = RC_TIMESTAMP_MODE }, 0, 0, VE, "rc_mode" },
92 static const AVClass class = {
93 .class_name = "libopenh264enc",
94 .item_name = av_default_item_name,
96 .version = LIBAVUTIL_VERSION_INT,
99 static av_cold int svc_encode_close(AVCodecContext *avctx)
101 SVCContext *s = avctx->priv_data;
104 WelsDestroySVCEncoder(s->encoder);
106 av_log(avctx, AV_LOG_WARNING, "%d frames skipped\n", s->skipped);
110 static av_cold int svc_encode_init(AVCodecContext *avctx)
112 SVCContext *s = avctx->priv_data;
113 SEncParamExt param = { 0 };
116 WelsTraceCallback callback_function;
117 AVCPBProperties *props;
119 if ((err = ff_libopenh264_check_version(avctx)) < 0)
122 if (WelsCreateSVCEncoder(&s->encoder)) {
123 av_log(avctx, AV_LOG_ERROR, "Unable to create encoder\n");
124 return AVERROR_UNKNOWN;
127 // Pass all libopenh264 messages to our callback, to allow ourselves to filter them.
128 log_level = WELS_LOG_DETAIL;
129 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_LEVEL, &log_level);
131 // Set the logging callback function to one that uses av_log() (see implementation above).
132 callback_function = (WelsTraceCallback) ff_libopenh264_trace_callback;
133 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_CALLBACK, &callback_function);
135 // Set the AVCodecContext as the libopenh264 callback context so that it can be passed to av_log().
136 (*s->encoder)->SetOption(s->encoder, ENCODER_OPTION_TRACE_CALLBACK_CONTEXT, &avctx);
138 (*s->encoder)->GetDefaultParams(s->encoder, ¶m);
140 #if FF_API_CODER_TYPE
141 FF_DISABLE_DEPRECATION_WARNINGS
143 s->cabac = avctx->coder_type == FF_CODER_TYPE_AC;
144 FF_ENABLE_DEPRECATION_WARNINGS
147 param.fMaxFrameRate = 1/av_q2d(avctx->time_base);
148 param.iPicWidth = avctx->width;
149 param.iPicHeight = avctx->height;
150 param.iTargetBitrate = avctx->bit_rate > 0 ? avctx->bit_rate : TARGET_BITRATE_DEFAULT;
151 param.iMaxBitrate = FFMAX(avctx->rc_max_rate, avctx->bit_rate);
152 param.iRCMode = s->rc_mode;
153 if (avctx->qmax >= 0)
154 param.iMaxQp = av_clip(avctx->qmax, 1, 51);
155 if (avctx->qmin >= 0)
156 param.iMinQp = av_clip(avctx->qmin, 1, param.iMaxQp);
157 param.iTemporalLayerNum = 1;
158 param.iSpatialLayerNum = 1;
159 param.bEnableDenoise = 0;
160 param.bEnableBackgroundDetection = 1;
161 param.bEnableAdaptiveQuant = 1;
162 param.bEnableFrameSkip = s->skip_frames;
163 param.bEnableLongTermReference = 0;
164 param.iLtrMarkPeriod = 30;
165 if (avctx->gop_size >= 0)
166 param.uiIntraPeriod = avctx->gop_size;
167 #if OPENH264_VER_AT_LEAST(1, 4)
168 param.eSpsPpsIdStrategy = CONSTANT_ID;
170 param.bEnableSpsPpsIdAddition = 0;
172 param.bPrefixNalAddingCtrl = 0;
173 param.iLoopFilterDisableIdc = !s->loopfilter;
174 param.iEntropyCodingModeFlag = 0;
175 param.iMultipleThreadIdc = avctx->thread_count;
176 if (s->profile && !strcmp(s->profile, "main"))
177 param.iEntropyCodingModeFlag = 1;
178 else if (!s->profile && s->cabac)
179 param.iEntropyCodingModeFlag = 1;
181 param.sSpatialLayers[0].iVideoWidth = param.iPicWidth;
182 param.sSpatialLayers[0].iVideoHeight = param.iPicHeight;
183 param.sSpatialLayers[0].fFrameRate = param.fMaxFrameRate;
184 param.sSpatialLayers[0].iSpatialBitrate = param.iTargetBitrate;
185 param.sSpatialLayers[0].iMaxSpatialBitrate = param.iMaxBitrate;
187 #if OPENH264_VER_AT_LEAST(1, 7)
188 if (avctx->sample_aspect_ratio.num && avctx->sample_aspect_ratio.den) {
190 static const AVRational sar_idc[] = {
191 { 0, 0 }, // Unspecified (never written here).
192 { 1, 1 }, { 12, 11 }, { 10, 11 }, { 16, 11 },
193 { 40, 33 }, { 24, 11 }, { 20, 11 }, { 32, 11 },
194 { 80, 33 }, { 18, 11 }, { 15, 11 }, { 64, 33 },
195 { 160, 99 }, // Last 3 are unknown to openh264: { 4, 3 }, { 3, 2 }, { 2, 1 },
197 static const ESampleAspectRatio asp_idc[] = {
199 ASP_1x1, ASP_12x11, ASP_10x11, ASP_16x11,
200 ASP_40x33, ASP_24x11, ASP_20x11, ASP_32x11,
201 ASP_80x33, ASP_18x11, ASP_15x11, ASP_64x33,
206 av_reduce(&num, &den, avctx->sample_aspect_ratio.num,
207 avctx->sample_aspect_ratio.den, 65535);
209 for (i = 1; i < FF_ARRAY_ELEMS(sar_idc); i++) {
210 if (num == sar_idc[i].num &&
211 den == sar_idc[i].den)
214 if (i == FF_ARRAY_ELEMS(sar_idc)) {
215 param.sSpatialLayers[0].eAspectRatio = ASP_EXT_SAR;
216 param.sSpatialLayers[0].sAspectRatioExtWidth = num;
217 param.sSpatialLayers[0].sAspectRatioExtHeight = den;
219 param.sSpatialLayers[0].eAspectRatio = asp_idc[i];
221 param.sSpatialLayers[0].bAspectRatioPresent = true;
223 param.sSpatialLayers[0].bAspectRatioPresent = false;
227 if ((avctx->slices > 1) && (s->max_nal_size)) {
228 av_log(avctx, AV_LOG_ERROR,
229 "Invalid combination -slices %d and -max_nal_size %d.\n",
230 avctx->slices, s->max_nal_size);
231 return AVERROR(EINVAL);
234 if (avctx->slices > 1)
235 s->slice_mode = SM_FIXEDSLCNUM_SLICE;
238 s->slice_mode = SM_SIZELIMITED_SLICE;
240 #if OPENH264_VER_AT_LEAST(1, 6)
241 param.sSpatialLayers[0].sSliceArgument.uiSliceMode = s->slice_mode;
242 param.sSpatialLayers[0].sSliceArgument.uiSliceNum = avctx->slices;
244 param.sSpatialLayers[0].sSliceCfg.uiSliceMode = s->slice_mode;
245 param.sSpatialLayers[0].sSliceCfg.sSliceArgument.uiSliceNum = avctx->slices;
248 if (s->slice_mode == SM_SIZELIMITED_SLICE) {
249 if (s->max_nal_size) {
250 param.uiMaxNalSize = s->max_nal_size;
251 #if OPENH264_VER_AT_LEAST(1, 6)
252 param.sSpatialLayers[0].sSliceArgument.uiSliceSizeConstraint = s->max_nal_size;
254 param.sSpatialLayers[0].sSliceCfg.sSliceArgument.uiSliceSizeConstraint = s->max_nal_size;
257 av_log(avctx, AV_LOG_ERROR, "Invalid -max_nal_size, "
258 "specify a valid max_nal_size to use -slice_mode dyn\n");
259 return AVERROR(EINVAL);
263 if ((*s->encoder)->InitializeExt(s->encoder, ¶m) != cmResultSuccess) {
264 av_log(avctx, AV_LOG_ERROR, "Initialize failed\n");
265 return AVERROR_UNKNOWN;
268 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
269 SFrameBSInfo fbi = { 0 };
271 (*s->encoder)->EncodeParameterSets(s->encoder, &fbi);
272 for (i = 0; i < fbi.sLayerInfo[0].iNalCount; i++)
273 size += fbi.sLayerInfo[0].pNalLengthInByte[i];
274 avctx->extradata = av_mallocz(size + AV_INPUT_BUFFER_PADDING_SIZE);
275 if (!avctx->extradata)
276 return AVERROR(ENOMEM);
277 avctx->extradata_size = size;
278 memcpy(avctx->extradata, fbi.sLayerInfo[0].pBsBuf, size);
281 props = ff_add_cpb_side_data(avctx);
283 return AVERROR(ENOMEM);
284 props->max_bitrate = param.iMaxBitrate;
285 props->avg_bitrate = param.iTargetBitrate;
290 static int svc_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
291 const AVFrame *frame, int *got_packet)
293 SVCContext *s = avctx->priv_data;
294 SFrameBSInfo fbi = { 0 };
297 SSourcePicture sp = { 0 };
298 int size = 0, layer, first_layer = 0;
299 int layer_size[MAX_LAYER_NUM_OF_FRAME] = { 0 };
301 sp.iColorFormat = videoFormatI420;
302 for (i = 0; i < 3; i++) {
303 sp.iStride[i] = frame->linesize[i];
304 sp.pData[i] = frame->data[i];
306 sp.iPicWidth = avctx->width;
307 sp.iPicHeight = avctx->height;
309 if (frame->pict_type == AV_PICTURE_TYPE_I) {
310 (*s->encoder)->ForceIntraFrame(s->encoder, true);
313 encoded = (*s->encoder)->EncodeFrame(s->encoder, &sp, &fbi);
314 if (encoded != cmResultSuccess) {
315 av_log(avctx, AV_LOG_ERROR, "EncodeFrame failed\n");
316 return AVERROR_UNKNOWN;
318 if (fbi.eFrameType == videoFrameTypeSkip) {
320 av_log(avctx, AV_LOG_DEBUG, "frame skipped\n");
324 // Normal frames are returned with one single layer, while IDR
325 // frames have two layers, where the first layer contains the SPS/PPS.
326 // If using global headers, don't include the SPS/PPS in the returned
327 // packet - thus, only return one layer.
328 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER)
329 first_layer = fbi.iLayerNum - 1;
331 for (layer = first_layer; layer < fbi.iLayerNum; layer++) {
332 for (i = 0; i < fbi.sLayerInfo[layer].iNalCount; i++)
333 layer_size[layer] += fbi.sLayerInfo[layer].pNalLengthInByte[i];
334 size += layer_size[layer];
336 av_log(avctx, AV_LOG_DEBUG, "%d slices\n", fbi.sLayerInfo[fbi.iLayerNum - 1].iNalCount);
338 if ((ret = ff_alloc_packet2(avctx, avpkt, size, size))) {
339 av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
343 for (layer = first_layer; layer < fbi.iLayerNum; layer++) {
344 memcpy(avpkt->data + size, fbi.sLayerInfo[layer].pBsBuf, layer_size[layer]);
345 size += layer_size[layer];
347 avpkt->pts = frame->pts;
348 if (fbi.eFrameType == videoFrameTypeIDR)
349 avpkt->flags |= AV_PKT_FLAG_KEY;
354 static const AVCodecDefault svc_enc_defaults[] = {
362 AVCodec ff_libopenh264_encoder = {
363 .name = "libopenh264",
364 .long_name = NULL_IF_CONFIG_SMALL("OpenH264 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
365 .type = AVMEDIA_TYPE_VIDEO,
366 .id = AV_CODEC_ID_H264,
367 .priv_data_size = sizeof(SVCContext),
368 .init = svc_encode_init,
369 .encode2 = svc_encode_frame,
370 .close = svc_encode_close,
371 .capabilities = AV_CODEC_CAP_AUTO_THREADS,
372 .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE | FF_CODEC_CAP_INIT_CLEANUP,
373 .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P,
375 .defaults = svc_enc_defaults,
376 .priv_class = &class,
377 .wrapper_name = "libopenh264",