So I'm writing an app that blurs and downscales large quantities of images, but all the image processing libraries I've come across define gaussian blurs in terms of sigma and kernel size rather than pixel radius (like an image editing app would do). Does radius relate to sigma and kernel size in some way, or is it an entirely different thing?
1 Answer
After a bit more digging I found a gaussian kernel calculator that seems to suggest sigma is roughly half the blur radius, and kernel size should be somewhere around (sigma*4)+1 for a full blur.