不久前刚实现SkeyeRTMPPusher扩展支持h265推送,当时在网上也查找了很多资料,发现都不尽详细,而官方也没有更新对HEVC(H265,后文统称HEVC)tag的支持,反正是走了不少弯路,当然,在广大网友以及ffmpeg代码的帮助下我最终实现了通过SkeyeRTMPPusher推送HEVC视频帧数据到SkeyeSMS,这里我将把实现过程详细的记录下来,供广大网友参考。
代码语言:txt复制首先, RTMP头部信息封装并没有定义HEVC,我们采用CDN联盟的HEVC扩展标准,将HEVC的VideoTagHeader定义为12,详见下图:
代码语言:txt复制然后,我们在H264封装的基础上进行改进,以支持HEVC头部的封装,而HEVC头有
SPS PPS VPS,我们参考ffmpeg的HEVCDecoderConfigurationRecord结构对metadata进行封装,该结构体代码如下:
代码语言:txt复制typedef struct HVCCNALUnitArray {
uint8_t array_completeness;
uint8_t NAL_unit_type;
uint16_t numNalus;
uint16_t *nalUnitLength;
uint8_t **nalUnit;
} HVCCNALUnitArray;
typedef struct HEVCDecoderConfigurationRecord {
uint8_t configurationVersion;
uint8_t general_profile_space;
uint8_t general_tier_flag;
uint8_t general_profile_idc;
uint32_t general_profile_compatibility_flags;
uint64_t general_constraint_indicator_flags;
uint8_t general_level_idc;
uint16_t min_spatial_segmentation_idc;
uint8_t parallelismType;
uint8_t chromaFormat;
uint8_t bitDepthLumaMinus8;
uint8_t bitDepthChromaMinus8;
uint16_t avgFrameRate;
uint8_t constantFrameRate;
uint8_t numTemporalLayers;
uint8_t temporalIdNested;
uint8_t lengthSizeMinusOne;
uint8_t numOfArrays;
HVCCNALUnitArray *array;
} HEVCDecoderConfigurationRecord;
参考ffmeg对该结构进行初始化如下:
代码语言:txt复制static void hvcc_init(HEVCDecoderConfigurationRecord *hvcc)
{
memset(hvcc, 0, sizeof(HEVCDecoderConfigurationRecord));
hvcc->configurationVersion = 1;
hvcc->lengthSizeMinusOne = 3; // 4 bytes
/*
* The following fields have all their valid bits set by default,
* the ProfileTierLevel parsing code will unset them when needed.
*/
hvcc->general_profile_compatibility_flags = 0xffffffff;
hvcc->general_constraint_indicator_flags = 0xffffffffffff;
/*
* Initialize this field with an invalid value which can be used to detect
* whether we didn't see any VUI (in which case it should be reset to zero).
*/
hvcc->min_spatial_segmentation_idc = MAX_SPATIAL_SEGMENTATION 1;
}
需要注意的是,该结构其他参数我们其实可以不特别关心,我们只需要在HVCCNALUnitArray数组中把HEVC的VPS,SPS和PPS信息填入即可。
最后,填写好Metadata信息后,我们还需要在发送帧数据的时候做一下修改,将I帧的tag头改成12,P帧tag不变,设置成1即可,如下代码所示:
代码语言:txt复制 int i = 0;
if(bIsKeyFrame)
{
//body[i ] = 0x17;// 2:Pframe 7:AVC
body[i ] = (m_metadata.nVideoCodec == FLV_CODECID_HEVC) ? 0x1C:0x17;// 1:Iframe 7:AVC 12:HEVC
if (m_bWaitingKeyFrame)
{
m_bWaitingKeyFrame = false;
}
}
else
{
//body[i ] = 0x27;// 2:Pframe 7:AVC
body[i ] = (m_metadata.nVideoCodec == FLV_CODECID_HEVC) ? 0x2C:0x27;// 1:Iframe 7:AVC 12:HEVC
}
body[i ] = 0x01;// AVC NALU
body[i ] = 0x00;
body[i ] = 0x00;
body[i ] = 0x00;
// NALU size
body[i ] = size>>24;
body[i ] = size>>16;
body[i ] = size>>8;
body[i ] = size&0xff;