WillBrennan/BlurDetection2

easier normalization for image size

AruniRC opened this issue · 0 comments

Hi, great project. I think there is an easier way for you to normalize for the different image sizes.
Instead of resizing to a pre-defined image size (which would make small, in-focus images to become blurry due to the upsampling), you could simply divide the "variance of laplacian" focus measure by the image dimensions. I've tried this on a dataset of images and it worked at least as well as resizing, qualitatively.