Core graphics RGB data issue - iphone

Question

I am trying to to pixel-by-pixel image filters using Core Graphics (breaking a CGImage into unsigned integers using CFData)
When I try to create an imaged with the processed data, however, the resulting image comes out with significantly different colors.
I commented out the entire loop where I actually alter the pixels' rgb values and nothing changes, either.
When I initialize the UIImage I am using in the filter; I do a resize using drawInRect with UIGraphicsBeginContext(); on an image taken from the camera.
When I remove the resize step and set my image directly from the camera; the filters seem to work just fine. Here's the code where I initialize the image I am using (from inside didFinishPickingImage)
self.editingImage is a UIImageView and self.editingUIImage is a UIImage
-(void)imagePickerController:(UIImagePickerController *)picker
didFinishPickingImage : (UIImage *)image
editingInfo:(NSDictionary *)editingInfo
{
self.didAskForImage = YES;
UIGraphicsBeginImageContext(self.editingImage.frame.size);
float prop = image.size.width / image.size.height;
float left, top, width, height;
if(prop < 1){
height = self.editingImage.frame.size.height;
width = (height / image.size.height) * image.size.width;
left = (self.editingImage.frame.size.width - width)/2;
top = 0;
}else{
width = self.editingImage.frame.size.width;
height = (width / image.size.width) * image.size.height;
top = (self.editingImage.frame.size.height - height)/2;
left = 0;
}
[image drawInRect:CGRectMake(left, top, width, height)];
self.editingUIImage = UIGraphicsGetImageFromCurrentImageContext();
UIGraphicsEndImageContext();
self.editingImage.image = self.editingUIImage;
[self.contrastSlider addTarget:self action:#selector(doImageFilter:) forControlEvents:UIControlEventValueChanged];
[self.brightnessSlider addTarget:self action:#selector(doImageFilter:) forControlEvents:UIControlEventValueChanged];
[picker dismissModalViewControllerAnimated:YES];
picker = nil;
}
The resizes the image just the way I need it as far as position;
Here's the image filtering function, I've taken the actual loop contents out because they're irrelevant.
- (void) doImageFilter:(id)sender{
CGImageRef src = self.editingUIImage.CGImage;
CFDataRef dta;
dta = CGDataProviderCopyData(CGImageGetDataProvider(src));
UInt8 *pixData = (UInt8 *) CFDataGetBytePtr(dta);
int dtaLen = CFDataGetLength(dta);
for (int i = 0; i < dtaLen; i += 3) {
//the loop
}
CGContextRef ctx;
ctx = CGBitmapContextCreate(pixData, CGImageGetWidth(src), CGImageGetHeight(src), 8, CGImageGetBytesPerRow(src), CGImageGetColorSpace(src), kCGImageAlphaPremultipliedLast);
CGImageRef newCG = CGBitmapContextCreateImage(ctx);
UIImage *new = [UIImage imageWithCGImage:newCG];
CGContextRelease(ctx);
CFRelease(dta);
CGImageRelease(newCG);
self.editingImage.image = new;
}
The image looks like this at first
and then after doing doImageFilter...
As mentioned before, this only happens when I use the resize method shown above.
Really stumped on this one, been researching it all day... any help very appreciated!
Cheers
Update: I've examined all the image objects' color spaces and they're all kCGColorSpaceDeviceRGB. Pretty stumped on this one guys, I'm pretty some something is going wrong when I break the image into unsigned integers, but I'm not sure what.. Anyone?

Answer 1

Your problem is on the last line:
ctx = CGBitmapContextCreate(pixData,
CGImageGetWidth(src),
CGImageGetHeight(src),
8,
CGImageGetBytesPerRow(src),
CGImageGetColorSpace(src),
kCGImageAlphaPremultipliedLast);
You're making an assumption about the alpha and the component ordering of the data of the source image, which is apparently not correct. You should get that from the source image via CGImageGetBitmapInfo(src).
To avoid issues like this one, if you're starting with an arbitrary CGImage and you want to manipulate the bytes of the bitmap directly, it is best to make a CGBitmapContext in a format that you specify yourself (not directly taken from the source image). Then, draw your source image into the bitmap context; CG will convert the image's data into your bitmap context's format, if necessary. Then get the data from the bitmap context and manipulate it.