Hello everyone, I have a bunch of images I would like to sort automatically.
These are screenshots from a videogames, so there are parts of the images that never change.
Is there a simple/quick way to sort those images, based on rules such as "if there is this bunch of pixels in the right hand corner, put in folder N#1", "if the middle of the bottom is composed of such and such pixels, put in folder N#2", and so on.
If you are sure that certain parts of the screen will always be the same and you want to look for those, just create a mask image (ie cut out the corner that you will compare against) and subtract it from all your images. Then you can filter those that pass the test based on the amount of pixels left in your mask region below some threshold value (in case of subtle error or alpha blending).
If you have an even simpler case where you can rely on pixels being in a certain area (say from a HUD) then you can just sample the areas you're interested in and bin the images based on the color sampled.
If your case is more complex (or the color values change significantly enough that you cannot generate a mask image that will reliably difference to sensible values) you're going to have to start looking at color correction, image registration, and object recognition; all of which are much more complex than you probably desire for this need.
Do you have a language/target platform in mind?