While we have been developing a website that displayed user uploaded images in fixed size regions, we have encountered some interesting problems. And this have led to some questions: Should we resize the image to fit the fixed size region ? Should we crop the image ? Maybe we should mix the two ?
After some tests we found that resizing the image is not the optimal solution, since the fixed size region is quite small. Then I tested the second and the third solution by cropping/resizing manually the images. The tests were convincing !
Needless to say we are not going to crop manually all the images ! we should automate this task. The no-brainer solution is to crop the image by focusing on the center. This naîve solution gave amazingly acceptable results, but sometimes it fails badly if the interesting region lies outside the cropped zone. Our algorithm should be able to somehow locate automatically the interesting part of the image.