How is the image data interpreted for a grayscale image on an iPhone?

https://stackoverflow.com/questions/4433969

09-10-2019
|

Question

How do I make sense of the image data for a grayscale image given the following scenario: I capture video data from the "sample buffer" and extract an 80x20 section and then turn that into a grayscale UIImage. But when I examine the raw pixel bytes I am unable to make sense of them in a way that would allow me to go on and "binarize" them (my real goal).

When I simply save the UIImage to the photo album using UIImageWriteToSavedPhotosAlbum to verify just what kind of image data I have, I indeed get a plain, white 80x20 image (it's actually light-grayish). I captured a plain white image to simplify things, expecting to see only values between, say, 200 or so and 255, and yet there are sections of the image data full of zeroes, that clearly suggest rows of black pixels. Any help is appreciated. The relevant code and the image data (16 pixels at a time) are below.

Here is how I create the 80x20 grayscale image from a portion of the CMSampleBufferRef video data:

UIImage *imageFromImage(UIImage *image, CGRect rect)
{   
    CGImageRef sourceImageRef = [image CGImage];  
    CGImageRef newImageRef = CGImageCreateWithImageInRect(sourceImageRef, rect);  

    CGImageRef grayScaleImg = grayscaleCGImageFromCGImage(newImageRef);
    CGImageRelease(newImageRef);  

    UIImage *newImage = [UIImage imageWithCGImage:grayScaleImg scale:1.0 orientation:UIImageOrientationLeft]; 

    return newImage;  
}  

CGImageRef grayscaleCGImageFromCGImage(CGImageRef inputImage) 
{
    size_t width = CGImageGetWidth(inputImage);
    size_t height = CGImageGetHeight(inputImage);

    // Create a gray scale context and render the input image into that
    CGColorSpaceRef colorspace = CGColorSpaceCreateDeviceGray();
    CGContextRef context = CGBitmapContextCreate(NULL, width, height, 8, 
                    4*width, colorspace, kCGBitmapByteOrderDefault);

    CGContextDrawImage(context, CGRectMake(0,0, width,height), inputImage);

    // Get an image representation of the grayscale context which the input
    //    was rendered into.
    CGImageRef outputImage = CGBitmapContextCreateImage(context);

    // Cleanup
    CGContextRelease(context);
    CGColorSpaceRelease(colorspace);

    return (CGImageRef)[(id)outputImage autorelease];
}

and then, when I use the following code to dump the pixel data to the Console:

    CGImageRef inputImage = [imgIn CGImage];
CGDataProviderRef dataProvider = CGImageGetDataProvider(inputImage);
CFDataRef imageData = CGDataProviderCopyData(dataProvider);
const UInt8 *rawData = CFDataGetBytePtr(imageData);

size_t width = CGImageGetWidth(inputImage);
    size_t height = CGImageGetHeight(inputImage);

    size_t numPixels = height * width;
for (int i = 0; i < numPixels ; i++)
{   
   if ((i % 16) == 0)
          NSLog(@" -%i-%i-%i-%i-%i-%i-%i-%i-%i-%i-%i-%i-%i-%i-%i-%i-\n\n", rawData[i],         
             rawData[i+1], rawData[i+2], rawData[i+3], rawData[i+4], rawData[i+5], 
             rawData[i+6], rawData[i+7], rawData[i+8], rawData[i+9], rawData[i+10], 
             rawData[i+11], rawData[i+12], rawData[i+13], rawData[i+14], rawData[i+15]);
}

I consistently get output like following:

-216-217-214-215-217-215-216-213-214-214-214-215-215-217-216-216-

-219-219-216-219-220-217-212-214-215-214-217-220-219-217-214-219-

-216-216-218-217-218-221-217-213-214-212-214-212-212-214-214-213-

-213-213-212-213-212-214-216-214-212-210-211-210-213-210-213-208-

-212-208-208-210-206-207-206-207-210-205-206-208-209-210-210-207-

-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-

(this pattern repeats for the remaining bytes, 80 bytes of pixel data in the 200's, depending on lighting, followed by 240 bytes of zeros -- there's a total of 1600 bytes since the image is 80x20)

Solution

This:

CGContextRef context = CGBitmapContextCreate(NULL, width, height, 8, 
                4*width, colorspace, kCGBitmapByteOrderDefault);

Should be:

CGContextRef context = CGBitmapContextCreate(NULL, width, height, 8, 
                width, colorspace, kCGBitmapByteOrderDefault);

In other words, for an 8 bit gray image, the number of bytes per row is the same as the width.

OTHER TIPS

You've probably forgotten image stride - you're assuming that your images are stored as width*height but several systems store them as stride*height where stride > width. The zeros are padding that you should skip.

By the way, what do you mean "binarize" ? I guess you mean quantize to a less grey levels ?

Licensed under: CC-BY-SA with attribution

Not affiliated with StackOverflow