After looking at various antialiasing implementations, I found one that I thought was especially interesting and applicable to the topics we’ve discussed in class. The approach itself is called Fast Approximate Anti-Aliasing (FXAA). This approach can be thought of in two steps. First, we find the edges of an image, ie the primary location of undesirable jaggies. We then antialias the edges, instead of the entire image. I figure one approach would be to perform a Fourier transform, apply an edge filter, then antialias the edges?

katamarisun

Very often @Mitchzof for edge filters we don't even need to do a Fourier transform - we can simply split an edge in the low pass and high pass components by, for example, blurring the image with a Gaussian, and taking the difference between the Gaussian and the original image to get the high pass (Laplacian) components. We can then anti-alias just the high pass component, and re-add them.

After looking at various antialiasing implementations, I found one that I thought was especially interesting and applicable to the topics we’ve discussed in class. The approach itself is called Fast Approximate Anti-Aliasing (FXAA). This approach can be thought of in two steps. First, we find the edges of an image, ie the primary location of undesirable jaggies. We then antialias the edges, instead of the entire image. I figure one approach would be to perform a Fourier transform, apply an edge filter, then antialias the edges?

Very often @Mitchzof for edge filters we don't even need to do a Fourier transform - we can simply split an edge in the low pass and high pass components by, for example, blurring the image with a Gaussian, and taking the difference between the Gaussian and the original image to get the high pass (Laplacian) components. We can then anti-alias just the high pass component, and re-add them.