天天看点

Mac OS使用FFmpeg进行视频H264,H265编码

一.概述

本文将在Mac os系统上使用FFmpeg进行音视频的H264,H265编码。

使用FFmpeg版本为4.2。

二、编码器初始化

有两点需要注意的是:

1.设置

pCodecContext->flags |= AV_CODEC_FLAG_GLOBAL_HEADER;

的目的是可以通过

pCodecContext->extradata

pCodecContext->extradata_size

提取到返回PPS,SPS,VPS数据,适用于直播场景,注释中也写的很清楚

Place global headers in extradata instead of every keyframe.

如果没有设置则会在

AVPacket.data

中和其他数据一起返回,适用于直接写入文件。

2.设置

pPacket.flags |= AV_PKT_FLAG_KEY

的目的可以在编码后的

AVPacket

中识别出是否为关键帧。

int ret;
    enum AVCodecID codecID = AV_CODEC_ID_H264;
    if (!kUseH264Encode) {
        codecID = AV_CODEC_ID_HEVC;
    }
    
    pCodec = avcodec_find_encoder(codecID);
    
    pCodecContext = avcodec_alloc_context3(pCodec);
    pCodecContext->codec_type = AVMEDIA_TYPE_VIDEO;
    pCodecContext->pix_fmt = AV_PIX_FMT_YUV420P;
    pCodecContext->width = 1280;
    pCodecContext->height = 720;
    pCodecContext->time_base.num = 1;
    pCodecContext->time_base.den = 25;
    pCodecContext->bit_rate = 1000 * 1000;
    pCodecContext->qmin = 10;
    pCodecContext->qmax = 51;
    pCodecContext->gop_size = 25;
    pCodecContext->max_b_frames = 0;
   // pCodecContext->flags |= AV_CODEC_FLAG_GLOBAL_HEADER;
    
    AVDictionary *param = NULL;
    if (kUseH264Encode) {
        av_dict_set(&param, "preset", "slow", 0);
        av_dict_set(&param, "tune", "zerolatency", 0);
    }else{
        av_dict_set(&param, "preset", "ultrafast", 0);
        av_dict_set(&param, "tune", "zero-latency", 0);
    }

    if (avcodec_open2(pCodecContext, pCodec, &param)<0) {
        return;
    }
    
    pFrame = av_frame_alloc();
    pFrame->width = pCodecContext->width;
    pFrame->height = pCodecContext->height;
    pFrame->format = pCodecContext->pix_fmt;
    ret = av_frame_get_buffer(pFrame, 0);
    if (ret < 0) {
        printf("ret == %s\n", av_err2str(ret));
    }
    //初始化avpacket
    av_init_packet(&pPacket);
    pPacket.flags |= AV_PKT_FLAG_KEY;
           

三、编码

在Mac OS系统或者iOS系统中,采集到的一般是

CMSampleBufferRef

对象,需要先从中拿到

CVPixelBufferRef

对象,再从其中提取

YUV

数据。而此处的

YUV

格式,是一种

two-plane

模式,即Y和UV分为两个Plane,但是UV(CbCr)为交错存储,而不是分为三个plane,需要最终转换为420P格式,即

YYYYUV

// 锁定imageBuffer内存地址开始进行编码
        if (CVPixelBufferLockBaseAddress(pixelBuffer, 0) == kCVReturnSuccess) {
            //获取Y分量的地址
            UInt8 *bufferPtr = (UInt8 *)CVPixelBufferGetBaseAddressOfPlane(pixelBuffer,0);
            //获取UV分量的地址
            UInt8 *bufferPtr1 = (UInt8 *)CVPixelBufferGetBaseAddressOfPlane(pixelBuffer,1);
            //根据像素获取图片的真实宽度&高度
            size_t width = CVPixelBufferGetWidth(pixelBuffer);
            size_t height = CVPixelBufferGetHeight(pixelBuffer);
            // 获取Y分量长度
            size_t bytesrow0 = CVPixelBufferGetBytesPerRowOfPlane(pixelBuffer,0);
            size_t bytesrow1  = CVPixelBufferGetBytesPerRowOfPlane(pixelBuffer,1);
            UInt8 *yuv420_data = (UInt8 *)malloc(width * height * 3 / 2);
            
            //将NV12数据转成YUV420P(I420)数据
            UInt8 *pY = bufferPtr;
            UInt8 *pUV = bufferPtr1;
            UInt8 *pU = yuv420_data + width * height;
            UInt8 *pV = pU + width * height / 4;
            for(int i =0;i<height;i++)
            {
                memcpy(yuv420_data+i*width,pY+i*bytesrow0,width);
            }
            for(int j = 0;j<height/2;j++)
            {
                for(int i =0;i<width/2;i++)
                {
                    *(pU++) = pUV[i<<1];
                    *(pV++) = pUV[(i<<1) + 1];
                }
                pUV += bytesrow1;
            }
            
            // 3.5.分别读取YUV的数据
            pFrame->data[0] = yuv420_data;
            pFrame->data[1] = pFrame->data[0] + width * height;
            pFrame->data[2] = pFrame->data[1] + (width * height) / 4;
            pFrame->pts = frameCount;
            
            // 5.对编码前的原始数据(AVFormat)利用编码器进行编码,将 pFrame 编码后的数据传入pkt 中
            int ret = avcodec_send_frame(pCodecContext, pFrame);
            if (ret != 0) {
                printf("Failed to encode! \n");
                CVPixelBufferUnlockBaseAddress(pixelBuffer, 0);
                return;
            }
            
            while (1) {
                ret = avcodec_receive_packet(pCodecContext, &pPacket);
                if (ret == AVERROR(EAGAIN) || ret == AVERROR_EOF)
                    break;
                else if (ret < 0) {
                    fprintf(stderr, "Error encoding audio frame\n");
                    break;
                }
                frameCount++;
                if (pPacket.flags & AV_PKT_FLAG_KEY) {
                    videoFrame.isKeyFrame =  YES;
                }

                //write file
           NSData *data = [NSData dataWithBytes:pPacket.data length:pPacket.size];
           if ([self.delegate respondsToSelector:@selector(videoEncoder:encodeData:)]) {
              [self.delegate videoEncoder:self encodeData:data];}
                //释放packet
                av_packet_unref(&pPacket);
            }
            
            // 7.释放yuv数据
            free(yuv420_data);
        }
        CVPixelBufferUnlockBaseAddress(pixelBuffer, 0);
}
           

四、提取SPS,PPS,VPS数据

上文说了要单独提取SPS,PPS,VPS数据,需开始设置

pCodecContext->flags |= AV_CODEC_FLAG_GLOBAL_HEADER

uint8_t *extra_data = pCodecContext->extradata;
int extra_size = pCodecContext->extradata_size;
           

1.H264编码时,拿到的

extra_data

如下所示:

00000001 6764001f acb300a0 0b742000 00030020 00000651 e3064d00 00000168 e9732c8b
           

很明显

SPS,PPS

被4个字节的

start code

=

00 00 00 01

分割开,

NALU header

只有一个字节:

00 00 00 01 67  ---> (0x67 & 0x1f) = 7 ---> PPS
00 00 00 01 68  ---> (0x68 & 0x1f) = 8 ---> SPS
           

代码如下:

int pos = 0;
int pps_pos = 0,pps_length = 0;
int sps_pos = 0,sps_length = 0;
while (pos < (extra_size - 4)) {
      if (extra_data[pos] == 0 &&
         extra_data[pos+1] == 0 &&
         extra_data[pos+2] == 0 &&
         extra_data[pos+3] == 1) {
        if ((extra_data[pos+4] & 0x1f) == 7) {//sps
            sps_pos = pos+4;
        }else if ((extra_data[pos+4] & 0x1f) == 8){//pps
             pps_pos = pos+4;
        }
     }
	pos ++;
}
sps_length = pps_pos - sps_pos - 4;
pps_length = extra_size - pps_pos;
           

2.H265编码时,同样方法拿到的

extra_data

提取

SPS,PPS,VPS``NALU header

有两个字节,提取方法如下:

00 00 00 01 40 01  ---> (0x40 & 0x7E)>>1 = 32 ---> VPS
00 00 00 01 42 01  ---> (0x42 & 0x7E)>>1 = 33 ---> SPS
00 00 00 01 44 01  ---> (0x44 & 0x7E)>>1 = 34 ---> PPS
           

需要注意的是,此处还可能包含被3个字节的

start code

=

00 00 01

分割开的

NAL_UNIT_SEI

数据:

00 00 01 4e 01  ---> (0x4e & 0x7E)>>1 = 39 ---> SEI
           

五、编码结束

编码结束时,需要冲洗编码器,将编码器中缓存的数据冲洗出来,防止丢帧。方法是发送

avcodec_send_frame(pCodecContext, NULL)

,当

avcodec_receive_packet

的返回值为

AVERROR_EOF

则表示冲洗完成。最后再释放内存。