[request] feature to produce match sensitivity values for a group of images

[Human]

I find that I often need to manually work with Sikuli's pattern match preview/setter feature to differentiate between two very similar images, and choose values that eliminate false positives and false negatives for each. This is a common behavior for image pairs like the enabled and disabled form of a button, and it can also come into play in other situations. There seems to be no heuristic that works all the time, since the actual sensitivity ends up being some combination of image size and how similar the two reference images really are.

Sikuli has the necessary tools under the hood to take reference images and determine if there exists a match sensitivity for each that will uniquely identify one image if both are present, but it would have to be coded, and the interface to the new functionality would have to be determined. One deviation from normal Sikuli behavior is that the match wouldn't take the current screen into account; it would have to either cobble together a fake screenshot consisting of the reference images, or do some other trick.

A sub-wishlist would be a way to tell the IDE, at image capture time, that you only want 1 match for the captured image. The IDE would then set the sensitivity for that image such that only one match for the newly-captured image is found, in the context of the screen in which it was captured. (Of course, it'd be up to the script writer to ensure that this threshold is used consistently in other parts of the script.)

[niknah (hankin0)]

I had this problem too with zoom in/out buttons. If you're searching for the exact image as it is on the screen look at the score of the match object like...
print find("....").score

It should be 1

If it's not 1, you may want to try out this small patch, I'm not sure if it made it into the last release cause it was posted a few weeks after that...
https://bugs.launchpad.net/sikuli/+bug/619253

Sikuli should return the best matches first normally.