Noise removal from image

Hello, How can know the type of noise in an image? And which type of filter can I use to remove the noise from this image? Finally, how can I be sure that the noise is really removed after applying the filter?

[Image Analyst]

Why do you think there's noise? Of course there is but if it's not a problem for whatever you need to do then you can just ignore it. What do you need to do? When you do this, what noise or issues in the image cause a problem with your operations or results?

[Sa rah]

In fact, I'll apply the DCT (Discrete Cosine Transform) on my images, and as we know, the DCT coefficients are divided into three bands: low, middle and high frequencies. The high frequencies represent noise and small variations(details). That's why I need to remove the noise, so that I have a better face recognition. I hope you understand what I mean here.

[Image Analyst]

That's what JPEG compression uses. So you're in luck - you can just use imwrite().

The problem is that low pass filtering to remove high frequencies removed both the noise and the details that are not noise. That's why spatial domain noise reduction methods usually work better, at least the more sophisticated ones do.

[Sa rah]

Then, can I use bilateral filtering, for example, since it can effectively remove both impulsive noise and additive noise simultaneously? Or there is a better one that I use?

Answer 1

[Image Analyst]

Last I heard, the best noise reduction filter is BM3D. That's what everyone seems to be trying to beat in papers published these days. I haven't seen anything beat it across the board. There are other good ones like non-local means, UINTA, K-LLD, K-SVD, etc. The bilateral filter is probably better than simple filters like median, averaging, etc. but the aforementioned ones are better, though a lot more complicated. Non-local means is in the File Exchange - maybe some of the others are too. MATLAB code for BM3D is on the link I gave you: http://www.cs.tut.fi/~foi/GCF-BM3D/