iOS使用VideoToolbox硬编码录制H264视频

如今各种直播如雨后春笋般在今年冒出来了，主要的移动平台Android和iOS都相继开放了视频硬件编解码接口，以往这些接口都是系统私有的，开发者无法使用。

视频编码解码是一项繁重的工作，尤其在移动平。所以从iOS8开始，苹果将VideoToolbox.framework开放了出来，使开发者可以使用iOS设备内置的硬件设备来进行视频的编码和解码工作。硬件编解码的好处是，复杂的计算由专门的硬件电路完成，往往比使用cpu计算更高效，速度更快，功耗也更低。

VideoToolbox 工作流程大致为:

• 创建Session ->设置编码相关参数 ->开始编码
• —> 循环输入源数据(yuv类型的数据，直接从摄像头获取)
• —> 获取编码后的h264数据
• 结束编码

以下为详细代码，内有详细注释，完整工程在文章末尾。

#pragma mark - videotoolbox methods
- (int)startEncodeSession:(int)width height:(int)height framerate:(int)fps bitrate:(int)bt
{
    OSStatus status;
    _frameCount = 0;

    VTCompressionOutputCallback cb = encodeOutputCallback;
    status = VTCompressionSessionCreate(kCFAllocatorDefault, width, height, kCMVideoCodecType_H264, NULL, NULL, NULL, cb, (__bridge void *)(self), &_encodeSesion);

    if (status != noErr) {
        NSLog(@"VTCompressionSessionCreate failed. ret=%d", (int)status);
        return -1;
    }

    // 设置实时编码输出，降低编码延迟
    status = VTSessionSetProperty(_encodeSesion, kVTCompressionPropertyKey_RealTime, kCFBooleanTrue);
    NSLog(@"set realtime  return: %d", (int)status);

    // h264 profile, 直播一般使用baseline，可减少由于b帧带来的延时
    status = VTSessionSetProperty(_encodeSesion, kVTCompressionPropertyKey_ProfileLevel, kVTProfileLevel_H264_Baseline_AutoLevel);
    NSLog(@"set profile   return: %d", (int)status);

    // 设置编码码率(比特率)，如果不设置，默认将会以很低的码率编码，导致编码出来的视频很模糊
    status  = VTSessionSetProperty(_encodeSesion, kVTCompressionPropertyKey_AverageBitRate, (__bridge CFTypeRef)@(bt)); // bps
    status += VTSessionSetProperty(_encodeSesion, kVTCompressionPropertyKey_DataRateLimits, (__bridge CFArrayRef)@[@(bt*2/8), @1]); // Bps
    NSLog(@"set bitrate   return: %d", (int)status);

    // 设置关键帧间隔，即gop size
    status = VTSessionSetProperty(_encodeSesion, kVTCompressionPropertyKey_MaxKeyFrameInterval, (__bridge CFTypeRef)@(fps*2));

    // 设置帧率，只用于初始化session，不是实际FPS
    status = VTSessionSetProperty(_encodeSesion, kVTCompressionPropertyKey_ExpectedFrameRate, (__bridge CFTypeRef)@(fps));
    NSLog(@"set framerate return: %d", (int)status);

    // 开始编码
    status = VTCompressionSessionPrepareToEncodeFrames(_encodeSesion);
    NSLog(@"start encode  return: %d", (int)status);

    return 0;
}


// 编码一帧图像，使用queue，防止阻塞系统摄像头采集线程
- (void) encodeFrame:(CMSampleBufferRef )sampleBuffer
{
    dispatch_sync(_encodeQueue, ^{
        CVImageBufferRef imageBuffer = (CVImageBufferRef)CMSampleBufferGetImageBuffer(sampleBuffer);

        // pts,必须设置，否则会导致编码出来的数据非常大，原因未知
        CMTime pts = CMTimeMake(_frameCount, 1000);
        CMTime duration = kCMTimeInvalid;

        VTEncodeInfoFlags flags;

        // 送入编码器编码
        OSStatus statusCode = VTCompressionSessionEncodeFrame(_encodeSesion,
                                                              imageBuffer,
                                                              pts, duration,
                                                              NULL, NULL, &flags);

        if (statusCode != noErr) {
            NSLog(@"H264: VTCompressionSessionEncodeFrame failed with %d", (int)statusCode);

            [self stopEncodeSession];
            return;
        }
    });
}

- (void) stopEncodeSession
{
    VTCompressionSessionCompleteFrames(_encodeSesion, kCMTimeInvalid);

    VTCompressionSessionInvalidate(_encodeSesion);

    CFRelease(_encodeSesion);
    _encodeSesion = NULL;
}

// 编码回调，每当系统编码完一帧之后，会异步掉用该方法，此为c语言方法
void encodeOutputCallback(void *userData, void *sourceFrameRefCon, OSStatus status, VTEncodeInfoFlags infoFlags,
                       CMSampleBufferRef sampleBuffer )
{
    if (status != noErr) {
        NSLog(@"didCompressH264 error: with status %d, infoFlags %d", (int)status, (int)infoFlags);
        return;
    }
    if (!CMSampleBufferDataIsReady(sampleBuffer))
    {
        NSLog(@"didCompressH264 data is not ready ");
        return;
    }
    ViewController* vc = (__bridge ViewController*)userData;

    // 判断当前帧是否为关键帧
    bool keyframe = !CFDictionaryContainsKey( (CFArrayGetValueAtIndex(CMSampleBufferGetSampleAttachmentsArray(sampleBuffer, true), 0)), kCMSampleAttachmentKey_NotSync);

    // 获取sps & pps数据. sps pps只需获取一次，保存在h264文件开头即可
    if (keyframe && !vc->_spsppsFound)
    {
        size_t spsSize, spsCount;
        size_t ppsSize, ppsCount;

        const uint8_t *spsData, *ppsData;

        CMFormatDescriptionRef formatDesc = CMSampleBufferGetFormatDescription(sampleBuffer);
        OSStatus err0 = CMVideoFormatDescriptionGetH264ParameterSetAtIndex(formatDesc, 0, &spsData, &spsSize, &spsCount, 0 );
        OSStatus err1 = CMVideoFormatDescriptionGetH264ParameterSetAtIndex(formatDesc, 1, &ppsData, &ppsSize, &ppsCount, 0 );

        if (err0==noErr && err1==noErr)
        {
            vc->_spsppsFound = 1;
            [vc writeH264Data:(void *)spsData length:spsSize addStartCode:YES];
            [vc writeH264Data:(void *)ppsData length:ppsSize addStartCode:YES];

            NSLog(@"got sps/pps data. Length: sps=%zu, pps=%zu", spsSize, ppsSize);
        }
    }

    size_t lengthAtOffset, totalLength;
    char *data;

    CMBlockBufferRef dataBuffer = CMSampleBufferGetDataBuffer(sampleBuffer);
    OSStatus error = CMBlockBufferGetDataPointer(dataBuffer, 0, &lengthAtOffset, &totalLength, &data);

    if (error == noErr) {
        size_t offset = 0;
        const int lengthInfoSize = 4; // 返回的nalu数据前四个字节不是0001的startcode，而是大端模式的帧长度length

        // 循环获取nalu数据
        while (offset < totalLength - lengthInfoSize) {
            uint32_t naluLength = 0;
            memcpy(&naluLength, data + offset, lengthInfoSize); // 获取nalu的长度，

            // 大端模式转化为系统端模式
            naluLength = CFSwapInt32BigToHost(naluLength);
            NSLog(@"got nalu data, length=%d, totalLength=%zu", naluLength, totalLength);

            // 保存nalu数据到文件
            [vc writeH264Data:data+offset+lengthInfoSize length:naluLength addStartCode:YES];

            // 读取下一个nalu，一次回调可能包含多个nalu
            offset += lengthInfoSize + naluLength;
        }
    }
}

// 保存h264数据到文件
- (void) writeH264Data:(void*)data length:(size_t)length addStartCode:(BOOL)b
{
    // 添加4字节的 h264 协议 start code
    const Byte bytes[] = "\x00\x00\x00\x01";

    if (_h264File) {
        if(b)
            fwrite(bytes, 1, 4, _h264File);

        fwrite(data, 1, length, _h264File);
    } else {
        NSLog(@"_h264File null error, check if it open successed");
    }
}

编码保存的h264数据保存到Document目录下，在Info.plist文件中添加Application supports iTunes file sharing ，并设置为YES，即可通过iTunes直接导出来，然后使用VLC播放器即可直接播放。

Demo 源码地址
WWDC关于VideoToolbox的视频: 百度云链接 密码: uzsr

==== 直播美颜相关问题，欢迎加群讨论：496010189 ====