Face Detection ios7 Coordinates Scaling Issue - ios

Question

I am using the Face Detection API and would like to know how to convert coordinates from large high resolution images to smaller images displayed on an UIImageView. So far, I have inverted the co-ordinate system of my image and container view so that it matches the Core Image coordinate system and I have also calculated the ratio of heights between my high resolution image and the dimensions of my image view, but the coordinates that I am getting are not accurate at all. I am assuming I cannot convert the points from the large image to the small image as easily as I thought. Can anyone please point out my mistake(s)?
[self.shownImageViewer setTransform:CGAffineTransformMakeScale(1,-1)];
[self.view setTransform:CGAffineTransformMakeScale(1,-1)];
//240 x 320
self.shownImageViewer.image = self.imageToShow;
yscale = 320/self.imageToShow.size.height;
xscale = 240/self.imageToShow.size.width;
height = 320;
CIImage *image = [[CIImage alloc] initWithCGImage:[self.imageToShow CGImage]];
CIContext *faceDetectionContext = [CIContext contextWithOptions:nil];
CIDetector *faceDetector = [CIDetector detectorOfType:CIDetectorTypeFace context:faceDetectionContext options:#{CIDetectorAccuracy:CIDetectorAccuracyHigh}];
NSArray * features = [faceDetector featuresInImage:image options:#{CIDetectorImageOrientation:[NSNumber numberWithInt:6]}];
for(CIFaceFeature *feature in features)
{
if(feature.hasLeftEyePosition)
self.leftEye = feature.leftEyePosition;
if(feature.hasRightEyePosition)
self.rightEye = feature.rightEyePosition;
if(feature.hasMouthPosition)
self.mouth = feature.mouthPosition;
}
NSLog(#"%g and %g",xscale*self.rightEye.x, yscale*self.rightEye.y);
NSLog(#"%g and %g",yscale*self.leftEye.x, yscale*self.leftEye.y);
NSLog(#"%g",height);
self.rightEyeMarker.center = CGPointMake(xscale*self.rightEye.x,yscale*self.rightEye.y);
self.leftEyeMarker.center = CGPointMake(xscale*self.leftEye.x,yscale*self.leftEye.y);

Answer 1

I would start by removing the transform from your image view. Just have the image view display the image in the orientation its in already. This will make the calculations a lot easier.
Now the CIFaceFeature outputs its features in image coordinates. But your imageView might be smaller or bigger. So first, keep it simple and setting the imageView's content mode to top left.
imageView.contentMode = UIViewContentModeTopLeft;
Now you dont have to scale the coordinates at all.
When you are happy with that set the contentMode to something more sensible like aspect fit.
imageView.contentMode = UIViewContentModeScaleAspectFit;
Now you need to scale the x and the y co-ordinates by multiplying each co-ordinate by the aspect fit ratio.
CGFloat xRatio = imageView.frame.size.width / image.size.width;
CGFloat yRatio = imageView.frame.size.height / image.size.height;
CGFloat aspectFitRatio = MIN(xRatio, yRatio);
Lastly you want to add the rotation back in. Try to avoid this if possible, e.g. fix you images so they are upright to begin with.