No replies to this topic

[rouncer]

say youve got a 16x16 1 bit image, and a store of a 16x16 1 bit image, and you want to get some rough estimate of how alike they are.

so you could just compare space and matter with space and matter, +1 for a union, -1 for a difference, then you could score the images this way.

just say you were comparing it to an image its incomplete recreation, and youd like to finish drawing the image by slowly adding correct pixels to it, and there are a few different pictures it was drawing at once.

how could you always make sure you append to the correct picture, by seeing it partially?

you see this is a form of inference, I was wondering, how cool do simple algorythms get that can roughly approximate a similarity test on perhaps an incomplete picture with the end result?

I was just wondering if anyone has a better method than the simple comparison i mentioned.
It doesnt have to be 100% correct (well, that would be image recognition) just some rougher estimated version.