AVCapture appendSampleBuffer

问题:

I am going insane with this one - have looked everywhere and tried anything and everything I can thinks of.

Am making an iPhone app that uses AVFoundation - specifically AVCapture to capture video using the iPhone camera.

I need to have a custom image that is overlayed on the video feed included in the recording.

So far I have the AVCapture session set up, can display the feed, access the frame, save it as a UIImage and marge the overlay Image onto it. Then convert this new UIImage into a CVPixelBufferRef. annnd to double check that the bufferRef is working I converted it back to a UIImage and it displays the image fine still.

The trouble starts when I try to convert the CVPixelBufferRef into a CMSampleBufferRef to append to the AVCaptureSessions assetWriterInput. The CMSampleBufferRef always returning NULL when I attempt to create it.

Here is the -(void)captureOutput function

- (void)captureOutput:(AVCaptureOutput *)captureOutput 
didOutputSampleBuffer:(CMSampleBufferRef)sampleBuffer 
    fromConnection:(AVCaptureConnection *)connection
{ 

 UIImage *botImage = [self imageFromSampleBuffer:sampleBuffer];
 UIImage *wheel = [self imageFromView:wheelView];

 UIImage *finalImage = [self overlaidImage:botImage :wheel];
 //[previewImage setImage:finalImage]; <- works -- the image is being merged into one UIImage

 CVPixelBufferRef pixelBuffer = NULL;
 CGImageRef cgImage = CGImageCreateCopy(finalImage.CGImage);
 CFDataRef image = CGDataProviderCopyData(CGImageGetDataProvider(cgImage));
 int status = CVPixelBufferCreateWithBytes(NULL,
             self.view.bounds.size.width,
             self.view.bounds.size.height,
             kCVPixelFormatType_32BGRA, 
             (void*)CFDataGetBytePtr(image), 
             CGImageGetBytesPerRow(cgImage), 
             NULL, 
             0,
             NULL, 
             &pixelBuffer);
 if(status == 0){
  OSStatus result = 0;

  CMVideoFormatDescriptionRef videoInfo = NULL;
  result = CMVideoFormatDescriptionCreateForImageBuffer(NULL, pixelBuffer, &videoInfo);
  NSParameterAssert(result == 0 && videoInfo != NULL);

  CMSampleBufferRef myBuffer = NULL;
  result = CMSampleBufferCreateForImageBuffer(kCFAllocatorDefault,
            pixelBuffer, true, NULL, NULL, videoInfo, NULL, &myBuffer);
  NSParameterAssert(result == 0 && myBuffer != NULL);//always null :S

  NSLog(@"Trying to append");

  if (!CMSampleBufferDataIsReady(myBuffer)){
   NSLog(@"sampleBuffer data is not ready");
   return;
  }

  if (![assetWriterInput isReadyForMoreMediaData]){
   NSLog(@"Not ready for data :(");
   return;
  }

  if (![assetWriterInput appendSampleBuffer:myBuffer]){   
   NSLog(@"Failed to append pixel buffer");
  }



 }

}

Another solution I keep hearing about is using a AVAssetWriterInputPixelBufferAdaptor which eliminates the need to do the messy CMSampleBufferRef wrapping. However I have scoured stacked and apple developer forums and docs and can't find a clear description or example on how to set this up or how to use it. If anyone has a working example of it could you please show me or help me nut out the above issue - have been working on this non-stop for a week and am at wits end.

Let me know if you need any other info

Thanks in advance,

Michael

回答1:

You need AVAssetWriterInputPixelBufferAdaptor, here is the code to create it :

    // Create dictionary for pixel buffer adaptor
NSDictionary *bufferAttributes = [NSDictionary dictionaryWithObjectsAndKeys:[NSNumber numberWithInt:kCVPixelFormatType_32BGRA], kCVPixelBufferPixelFormatTypeKey, nil];

// Create pixel buffer adaptor
m_pixelsBufferAdaptor = [[AVAssetWriterInputPixelBufferAdaptor alloc] initWithAssetWriterInput:assetWriterInput sourcePixelBufferAttributes:bufferAttributes];

And the code to use it :

// If ready to have more media data
if (m_pixelsBufferAdaptor.assetWriterInput.readyForMoreMediaData) {
    // Create a pixel buffer
    CVPixelBufferRef pixelsBuffer = NULL;
    CVPixelBufferPoolCreatePixelBuffer(NULL, m_pixelsBufferAdaptor.pixelBufferPool, &pixelsBuffer);

    // Lock pixel buffer address
    CVPixelBufferLockBaseAddress(pixelsBuffer, 0);

    // Create your function to set your pixels data in the buffer (in your case, fill with your finalImage data)
    [self yourFunctionToPutDataInPixelBuffer:CVPixelBufferGetBaseAddress(pixelsBuffer)];

    // Unlock pixel buffer address
    CVPixelBufferUnlockBaseAddress(pixelsBuffer, 0);

    // Append pixel buffer (calculate currentFrameTime with your needing, the most simplest way is to have a frame time starting at 0 and increment each time you write a frame with the time of a frame (inverse of your framerate))
    [m_pixelsBufferAdaptor appendPixelBuffer:pixelsBuffer withPresentationTime:currentFrameTime];

    // Release pixel buffer
    CVPixelBufferRelease(pixelsBuffer);
}

And don't forget to release your pixelsBufferAdaptor.

回答2:

I do it by using CMSampleBufferCreateForImageBuffer() .

OSStatus ret = 0;
CMSampleBufferRef sample = NULL;
CMVideoFormatDescriptionRef videoInfo = NULL;
CMSampleTimingInfo timingInfo = kCMTimingInfoInvalid;
timingInfo.presentationTimeStamp = pts;
timingInfo.duration = duration;

ret = CMVideoFormatDescriptionCreateForImageBuffer(NULL, pixel, &videoInfo);
if (ret != 0) {
    NSLog(@"CMVideoFormatDescriptionCreateForImageBuffer failed! %d", (int)ret);
    goto done;
}
ret = CMSampleBufferCreateForImageBuffer(kCFAllocatorDefault, pixel, true, NULL, NULL,
                                         videoInfo, &timingInfo, &sample);
if (ret != 0) {
    NSLog(@"CMSampleBufferCreateForImageBuffer failed! %d", (int)ret);
    goto done;
}