Gaussian Blur edges keep as a square instead of a getting rounded

Question

So I've implemented Gaussian Blur (1D convolution) by going over the pixels horizontally and then vertically on the processed image, to produce a final image. I've used the formula for Gauss Distribution that appears on wiki.

The problem is that when I'm trying to blur a small white square on a black background, its edges still appear like a square, instead of becoming rounder (as in Photoshop). Is it supposed to be like that? I've noticed that when I use the Gaussian Blur function several times with the same radius, this problem doesn't occur (it gets rounder), but then it becomes too slow.

Am I missing something here? Thanks.

Answer 1

Try to increase the variance in the gaussian, it will increase the blur because it will give more weight to the adjacent pixels.

Also you need to take at least 5 samples from the gaussian, a mask with only 3 samples will not blur enough