3

I am trying to reduce aliasing in an image using some filters. I tried the Gaussian variant of filters to remove the said high frequency patterns in the image, but I feel it's a bit too much of a blur.

I need a filter that can help me do some kind of anti-aliasing without doing too much blur.

I have searched a lot and unfortunately the results are not fruitful.

trichoplax is on Codidact now
  • 6,292
  • 3
  • 23
  • 69
Varun Vijaykumar
  • 203
  • 2
  • 7
  • What are you requirements? Gaussian kernels are designed to blur without artifacts. A different kernel that blurs but isn't Gaussian is the box filter. It will produce artifacts however. I suppose more importantly, why isn't Gaussian good enough? – ChaoSXDemon Jul 10 '17 at 20:26
  • 1
    What kind of aliasing is present? Is it a synthetic image with staircasing/jaggies? Is it a texture with spurious low-frequency components (like a high-frequency grating or checkerboard)? Is it a photo that's been badly upscaled? – Dan Hulme Jul 11 '17 at 11:07
  • The image consist of Jaggies mostly ... like for instance, i have an image filled with Grass and I want to make sure that the grass blades are not too blurred. – Varun Vijaykumar Jul 11 '17 at 11:22
  • a sinc filter can downsample "perfectly" such that any details that would cause aliasing are removed, while other details are left perfectly alone. https://en.wikipedia.org/wiki/Sinc_filter – Alan Wolfe Jul 15 '17 at 00:34

1 Answers1

0

Depending on the specific content of your input images, you could either use the Gaussian filter with kernel size varying with distance from the camera ( Link ) or use a slightly more complex bilateral filter, which is widely used to eliminate noise while keeping edges mostly intact.

Michal Kučera
  • 353
  • 1
  • 8