Can I median filter an image that is already noised with salt and pepper?

I want to imread and median filter an image which I downloaded and has already been added with salt and pepper from its source, not that I'm the one to add the noise myself and later remove it. Is this possible?

I tried the procedure explained in other pages by Imageanalyst by splitting the RGB channels and filter them individually before reconstructing them back using cat3. However, I still got back the noisy image without been filtered. I noticed that he added the noise himself in the code, his image wasn't originally and initially noised with salt and pepper which leads me to raising this question

Answer 1

[Image Analyst]

Then just skip the part where I created a noisy image and just start using your own. Attach your code if you still need help.

[Ayobami Meadows]

Yes I skipped the part you added the noise in your own but still got noisy image as final output. I've attached the code as well as the two noisy images I wanna use.

Please I also wanna ask if MSE and PSNR can be calculated for the restored image. If so, kindly give the code for doing that and how to plot their graphs. Thank you.

[Image Analyst]

That's not salt and pepper noise. It is isolated noise that is different from its surroundings but your noise has many values, not just 0 or 255. For example one noise pixel has a value of 134 but it is surrounded by pixels with values less than 50 so it looks noisy in that spot.

You'll have to use a different noise reduction algorithm. For example try to take the median filter and subtract it from the original. If any pixel is more than, say, 20 gray levels different, then replace that pixel. Something like

diffImage = abs(double(redMF) - double(redChannel));
noiseImage = diffImage > 20;

[Ayobami Meadows]

Everything I'm driving at here is to run two different programs entirely in two different matlab editors:

First program(first editor) is to generate a 0.05 salt and pepper noise and then save as jpeg file.

Second program(second editor) is to use median filter to filter the saved noisy image from first program. But here in this second program we are not generating any noise but rather using the salt and pepper noised image we saved in the other first program.

Since you said the noisy image I generated from the first program is NOT salt and pepper noise, I've attached the code I used in generating it as well as the output noisy image. Please tell me where I went wrong because it is salt and pepper noise I want to work it specifically.

[Image Analyst]

The imnoise() function gives true salt and pepper noise, as I thought.

Here is the problem, from your comments: "the generated noisy image of figure 2 is then saved as a jpg file from the plotted figure".

NEVER save an image in JPG format. It is a lossy compression format. Doing this changed the values from their pure 0 and 255 to something else. Again, do not use JPG format for image analysis. I, and I think much/most of the world, now uses PNG which is a lossless format. Use that and you will get the expected results just like I did in my demo.

[Ayobami Meadows]

Thanks Image Analyst. You absolutely solved my problem with your understandable explanations!