AVFoundation - Get grayscale image from Y plane (k

问题:

I'm using AVFoundation to take a video and I'm recording in kCVPixelFormatType_420YpCbCr8BiPlanarFullRange format. I want to make grayscale image directly from the Y plane of the YpCbCr format.

I've tried to create CGContextRef by calling CGBitmapContextCreate, but the problem is, that I don't know what colorspace and pixelformat to choose.

- (void)captureOutput:(AVCaptureOutput *)captureOutput 
didOutputSampleBuffer:(CMSampleBufferRef)sampleBuffer 
       fromConnection:(AVCaptureConnection *)connection 
{       
    CVImageBufferRef imageBuffer = CMSampleBufferGetImageBuffer(sampleBuffer); 
    CVPixelBufferLockBaseAddress(imageBuffer,0);        

    /* Get informations about the Y plane */
    uint8_t *YPlaneAddress = CVPixelBufferGetBaseAddressOfPlane(imageBuffer, 0);
    size_t bytesPerRow = CVPixelBufferGetBytesPerRowOfPlane(imageBuffer, 0);
    size_t width = CVPixelBufferGetWidthOfPlane(imageBuffer, 0);
    size_t height = CVPixelBufferGetHeightOfPlane(imageBuffer, 0);

    /* the problematic part of code */
    CGColorSpaceRef colorSpace = CGColorSpaceCreateDeviceGray();

    CGContextRef newContext = CGBitmapContextCreate(YPlaneAddress,
    width, height, 8, bytesPerRow, colorSpace, kCVPixelFormatType_1Monochrome);

    CGImageRef newImage = CGBitmapContextCreateImage(newContext); 
    UIImage *grayscaleImage = [[UIImage alloc] initWithCGImage:newImage];

    // process the grayscale image ... 
}

When I run the code above, I got this errors:

<Error>: CGBitmapContextCreateImage: invalid context 0x0
<Error>: CGBitmapContextCreate: unsupported parameter combination: 8 integer bits/component; 16 bits/pixel; 1-component color space; kCGImageAlphaPremultipliedLast; 192 bytes/row.

PS: Sorry for my english.

回答1:

If I'm not mistaken, you shouldn't go via a CGContext. Instead, you should create a data provider and then directly the image.

Another mistake in your code is the use of the kCVPixelFormatType_1Monochrome constant. It's a constant used in video processing (AV libraries), not in Core Graphics (CG libraries). Just use kCGImageAlphaNone. That a single component (gray) per pixel is needed (instead of three as for RGB) is derived from the color space.

It could look like this:

CGDataProviderRef  dataProvider = CGDataProviderCreateWithData(NULL, YPlaneAdress,
      height * bytesPerRow, NULL);
CGImageRef newImage = CGImageCreate(width, height, 8, 8, bytesPerRow,
      colorSpace, kCGImageAlphaNone, dataProvider, NULL, NO, kCGRenderingIntentDefault);
CGDataProviderRelease(dataProvider);