Measures performance of two approaches to comparing the content of UIImage
Small experiment to look at the performance of two different approaches to comparing the contents of UIImage instances.
My initial technique was to use CoreImage. My first step was to apply a difference blend to the two images and pass that output to CIAreaMaximum. If any of the red, green or blue channels of the final output were non-zero, that would indicate the images were different.
However, following a suggestion from Peter Steinberger, I ported this code to Swift to do a side by side comparision.
My Core Image solution, even without recreating a new CIContext with each compare, is a whole load slower! Looking at three 640x640 images 25 times gives these total timings on my iPad Pro:
fb_compareWithImage- execution time 0.156560003757477UIImageEqualToImage- execution time 3.25684899091721UIImageEqualToImage(2) - execution time 1.75236403942108
UIImageEqualToImage (2) - uses a single CIContext.
So, if you want to compare the contents of two images, stick with memcmp!