Salt and pepper noise removal

Hi

I have a set of thermographic data in the form of pixel-pixel-time (third order tensor). At some point, I want to remove the salt and pepper noise from the data. So I am planning to a median filter, in particular I want to use the command

medfilt2

J = medfilt2(I,[m n]) performs median filtering, where each output pixel contains the median value in the m-by-n neighborhood around the corresponding pixel in the input image.

https://it.mathworks.com/help/images/ref/medfilt2.html

What could be standard criteria to define the size m and n?

I am working with black and white (grayscale) images (which are my thermal images).

From other analysis I know the size (in pixels) of an area where I could expect a possible defect. Maybe this sizes could become my m and n?

Greetings

Luca

Answer 1

[Gayathri]

Hi @Luca Pecoriello,

I understand that you need to know how to select the values of hyperparameters, "m" and "n", while performing median filtering. These values are typically chosen through a process of trial and error. Some points to keep in mind while deciding the values would be

• Consider the level of detail in your images. Larger filter sizes can smooth out noise but may also blur important features. Balancing noise reduction and detail preservation is key.
• Ideally, the filter size should be smaller than the features you want to preserve to avoid altering them.
• Start with a small filter size, such as '3x3' or '5x5' and gradually increase until you achieve a satisfactory balance between noise reduction and detail preservation.

Hope this helps!