how to combine median and mean filter to remove noise at a time?

i need a hybrid method to combine median and mean filters thanks in advance.

[Image Analyst]

Now our answers don't make sense because you blew away the original question. You should have started a new question. To answer this new question, there are probably dozens of ways you could think up to do that. What algorithm are you specifically thinking of?

Answer 1

[Bjorn Gustavsson]

such a general question! One definition of noise I've seen is that it is all signal that you're not interested in. Defined thusly what is noise depends on what the observer is interested in - In my case looking at aurora borealis stars, comets are sparse point-like noise sources, while an astronomer might consider the aurora a diffuse background noise.

If you're doing an "academic project in an image" I suggest you need to have a standard text-book on image processing by your side for these type of questions. I suggest Gonzalez and Woods (and Eddins).

Answer 2

[Image Analyst]

Noise is anything that is not your signal. However it can even be your signal. Detection of photons is a Poisson stochastic process so if you don't have lots and lots of photons you will have noise that is the signal itself.

The median filter is common mostly because it is simple and one of the earliest filters used. It's not necessarily the best, and can blur your signal with more blur coming with larger window sizes. It's claim to fame is that it removes isolated impulsive noise without blurring true edges. However, like I said, it does do blurring as simple experimentation will reveal. There are tons of filters with more being published every month. No one can possibly keep track of all the various filters and their improvements so that's why the simple basic filters like median, averaging, etc. remain familiar and topmost in everyone's toolboxes. You're of course welcome to explore other ones (bilateral, mean shift, non-local means, BM3D, K-LLD, etc.) if you want to. These more sophisticated filters often/usually give better performance at the expense of a more complicated algorithm and more processing time.