I am working on anisotropic diffusion and the two coefficients proposed by Perona & Malik.

I wanted to know what is the use of diffusion in image processing? Why is anisotropic diffusion important and in which fields is it commonly used?


1 Answer 1


The anisotropic diffusion algorithm by Perona and Malik is the pioneering work in partial derivatives equations (PDE)-based denoising.

It applies the law of diffusion on pixel intensities to smooth textures in an image. A threshold function is used to prevent diffusion to happen across edges, and therefore it preserves edges in the image. (Unlike for instance gaussian blur filter.) This makes it very interesting if you want to remove noise, but do not want to smooth out the edges of your image, for instance if you want to use these edges to segment the image, without being perturbed by the noise.

Numerous efforts have been made that build on it, improving or extending it.

Now for where it is used, I only have a limited culture. I can cite two

  • Image analysis in the Life-Science field (where I work): the images that you can get out of a microscope are extremely noisy, and most of the time, it is by construction. Automated image analysis of these data often involve segmentation, for which you sometimes PDE-based algos.

  • Video games! For instance, try playing Mass Effect (the first one at least).

  • $\begingroup$ Would you comment on where it is used in video games? I'm doing analysis of video game footage (SSF4:AE, UMvC3) and am interested in how this might be applied to those analysis. $\endgroup$
    – casperOne
    Jan 5, 2012 at 4:49
  • 1
    $\begingroup$ @casperOne: I cannot actually, since I do not belong to the video games experts domain. I noticed it was used (or a similar algo) to give to image a kind of "cinematic" feeling to 3D engines (combination of speckle noise + AD). In the aforementioned Mass Effect example, you can notice it, and even see the iterative process, by focusing the camera on the character and remaining still. Maybe a question on a specialized stack exchange forum might help you? $\endgroup$
    – Jean-Yves
    Jan 6, 2012 at 7:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.