H265（HEVC） nal 單元頭介紹及rtp傳送中的fu分組傳送詳解
阿新 • • 發佈：2019-02-14
首先來介紹下h265（HEVC）nal單元頭，與h264的nal層相比，h265的nal unit header有兩個位元組構成，如下圖所示：
從圖中可以看出hHEVC的nal包結構與h264有明顯的不同，hevc加入了nal所在的時間層的ＩＤ，取去除了nal_ref_idc，此資訊合併到了naltype中，通常情況下F為0，layerid為0,TID為1。
nal單元的型別有如下幾種：
下面接收下fu分組打包方式，fu分組包頭格式如下：enum NalUnitType { NAL_UNIT_CODED_SLICE_TRAIL_N = 0, // 0 NAL_UNIT_CODED_SLICE_TRAIL_R, // 1 NAL_UNIT_CODED_SLICE_TSA_N, // 2 NAL_UNIT_CODED_SLICE_TLA, // 3 // Current name in the spec: TSA_R NAL_UNIT_CODED_SLICE_STSA_N, // 4 NAL_UNIT_CODED_SLICE_STSA_R, // 5 NAL_UNIT_CODED_SLICE_RADL_N, // 6 NAL_UNIT_CODED_SLICE_DLP, // 7 // Current name in the spec: RADL_R NAL_UNIT_CODED_SLICE_RASL_N, // 8 NAL_UNIT_CODED_SLICE_TFD, // 9 // Current name in the spec: RASL_R NAL_UNIT_RESERVED_10, NAL_UNIT_RESERVED_11, NAL_UNIT_RESERVED_12, NAL_UNIT_RESERVED_13, NAL_UNIT_RESERVED_14, NAL_UNIT_RESERVED_15, NAL_UNIT_CODED_SLICE_BLA, // 16 // Current name in the spec: BLA_W_LP NAL_UNIT_CODED_SLICE_BLA, // 16 // Current name in the spec: BLA_W_LP NAL_UNIT_CODED_SLICE_BLANT, // 17 // Current name in the spec: BLA_W_DLP NAL_UNIT_CODED_SLICE_BLA_N_LP, // 18 NAL_UNIT_CODED_SLICE_IDR, // 19 // Current name in the spec: IDR_W_DLP NAL_UNIT_CODED_SLICE_IDR_N_LP, // 20 NAL_UNIT_CODED_SLICE_CRA, // 21 NAL_UNIT_RESERVED_22, NAL_UNIT_RESERVED_23, NAL_UNIT_RESERVED_24, NAL_UNIT_RESERVED_25, NAL_UNIT_RESERVED_26, NAL_UNIT_RESERVED_27, NAL_UNIT_RESERVED_28, NAL_UNIT_RESERVED_29, NAL_UNIT_RESERVED_30, NAL_UNIT_RESERVED_31, NAL_UNIT_VPS, // 32 NAL_UNIT_SPS, // 33 NAL_UNIT_PPS, // 34 NAL_UNIT_ACCESS_UNIT_DELIMITER, // 35 NAL_UNIT_EOS, // 36 NAL_UNIT_EOB, // 37 NAL_UNIT_FILLER_DATA, // 38 NAL_UNIT_SEI, // 39 Prefix SEI NAL_UNIT_SEI_SUFFIX, // 40 Suffix SEI NAL_UNIT_RESERVED_41, NAL_UNIT_RESERVED_42, NAL_UNIT_RESERVED_43, NAL_UNIT_RESERVED_44, NAL_UNIT_RESERVED_45, NAL_UNIT_RESERVED_46, NAL_UNIT_RESERVED_47, NAL_UNIT_UNSPECIFIED_48, NAL_UNIT_UNSPECIFIED_49, NAL_UNIT_UNSPECIFIED_50, NAL_UNIT_UNSPECIFIED_51, NAL_UNIT_UNSPECIFIED_52, NAL_UNIT_UNSPECIFIED_53, NAL_UNIT_UNSPECIFIED_54, NAL_UNIT_UNSPECIFIED_55, NAL_UNIT_UNSPECIFIED_56, NAL_UNIT_UNSPECIFIED_57, NAL_UNIT_UNSPECIFIED_58, NAL_UNIT_UNSPECIFIED_59, NAL_UNIT_UNSPECIFIED_60, NAL_UNIT_UNSPECIFIED_61, NAL_UNIT_UNSPECIFIED_62, NAL_UNIT_UNSPECIFIED_63, NAL_UNIT_INVALID, };
fus包頭包含了兩個位元組的payloadhdr，一個位元組的fu header，fu header與h264一樣，結構如下圖，包含開始位（1b）、停止位(1b)、futype(6b)
paylodhdr兩個自己的賦值，其實就是把hevc幀資料的nal unit header的naltype替換為49即可，下面是從ffmpeg原始碼中截取出來的fu打包方式程式碼片段：
static void nal_send(AVFormatContext *ctx, const uint8_t *buf, int len, int last_packet_of_frame) { RTPMuxContext *rtp_ctx = ctx->priv_data; int rtp_payload_size = rtp_ctx->max_payload_size - RTP_HEVC_HEADERS_SIZE; int nal_type = (buf[0] >> 1) & 0x3F; /* send it as one single NAL unit? */ if (len <= rtp_ctx->max_payload_size) //小於對定的最大值時，直接傳送（最大值一般小於mtu） { /* use the original NAL unit buffer and transmit it as RTP payload */ ff_rtp_send_data(ctx, buf, len, last_packet_of_frame); } else //大於最大值時進行fu分組傳送 { /* create the HEVC payload header and transmit the buffer as fragmentation units (FU) 0 1 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |F| Type | LayerId | TID | +-------------+-----------------+ F = 0 Type = 49 (fragmentation unit (FU)) LayerId = 0 TID = 1 */ rtp_ctx->buf[0] = 49 << 1; rtp_ctx->buf[1] = 1; //此處為paylaodhdr,規範賦值應該是替換hevc資料nal 的payloadhdr的type //rtp_ctx->buf[0] = (buf[0] &0x81) | (49<<1); //rtp_ctx->buf[1] = buf[1] /* create the FU header 0 1 2 3 4 5 6 7 +-+-+-+-+-+-+-+-+ |S|E| FuType | +---------------+ S = variable E = variable FuType = NAL unit type */ rtp_ctx->buf[2] = nal_type; /* set the S bit: mark as start fragment */ rtp_ctx->buf[2] |= 1 << 7; /* pass the original NAL header */ //此處要注意，當是分組的第一報資料時，應該覆蓋掉前兩個位元組的資料，h264要覆蓋前一個位元組的資料，即是第一包要去除hevc幀資料的paylaodhdr buf += 2; len -= 2; while (len > rtp_payload_size) { /* complete and send current RTP packet */ memcpy(&rtp_ctx->buf[RTP_HEVC_HEADERS_SIZE], buf, rtp_payload_size); ff_rtp_send_data(ctx, rtp_ctx->buf, rtp_ctx->max_payload_size, 0); buf += rtp_payload_size; len -= rtp_payload_size; /* reset the S bit */ rtp_ctx->buf[2] &= ~(1 << 7); } /* set the E bit: mark as last fragment */ rtp_ctx->buf[2] |= 1 << 6; /* complete and send last RTP packet */ memcpy(&rtp_ctx->buf[RTP_HEVC_HEADERS_SIZE], buf, len); ff_rtp_send_data(ctx, rtp_ctx->buf, len + 2, last_packet_of_frame); } }
通過rtp傳送hevc視訊資料，當hevc幀資料大於mtu時，應該進行fu分組傳送，從上面程式碼流程就是對超過max_payload_size資料進行fu分組的流程，這個h264 fu-A很類似，很容易理解。
參考規範：
https://tools.ietf.org/html/draft-ietf-payload-rtp-h265-14
ffmpeg相關程式碼
https://www.ffmpeg.org/doxygen/2.5/rtpenc__hevc_8c_source.html