How to remove noise without effecting the image

hi,

I have been trying to do image enhancement, but I found the noise is there any way to remove such kind of noise and what to call this kind of Noise given in following image

I tried median filter but it blur this image, kind of lost some details in image by adding blurring effect. Thanks for your Suggestions

[Walter Roberson]

Which noise? The second member of the pair has a portion enclosed by a black oval-like shape, but I do not notice any particular noise being pointed out.

[Muhammad Ali Qadar]

Walter Roberson, if you look closely in the second picture at the tire portion there is scattered black patches of noise first one is original image and second is enhanced one.

Answer 1

[Image Analyst]

What kind of noise is it? All noise removal methods will affect the image - they have to otherwise the image won't change and get noise removed. Why do you need that noise removed? It doesn't look that bad - bad enough to prevent you from doing whatever else you need to do with the image.

There are many types of noise removal and many are best suited for a particular kind of noise. If you have isolated salt and pepper noise, I have a modified median filter that does a great job on that. There are other denoising methods with different properties and effect. It seems BM3D is the current state of the art that people are trying to beat. There are others that are almost as good and perhaps even better for certain types of images. They range from the simple mean and median to slightly more complicated sigma filters and knn filters to more sophisticated and complicated mean shift, non-local means, BM3D, K-LLD, UINTA, K-SVD, plus more being invented every month. Look up some of those. I know there are some non-local means programs in the File Exchange. Maybe you can start with those.

[Image Analyst]

Muhammad Ali Qadar's "Answer" moved here:

I am segmenting the image using threshold values, the image would be removed based on best threshold values. So just using multithresh would not result into good threshold values. There are other methods like mean , median based or just apply clipping process

[Image Analyst]

I don't know what that means. What do you mean by "the image would be removed"? Do you mean the noise would be removed? Or that you would segment the image? Or that you would crop the image?

And who said anything about multithresh?

Yes, I mentioned mean and median, plus several other better methods.

And I don't know what you mean by "the clipping process."

[Muhammad Ali Qadar]

Sorry for my Sentence its like this "image noise would be removed". infact I am working on Image enhancement methods based on Brightness preservation of images. So There are methods that Partition the histogram or clip the histogram to improve the visual quality of image. there are methods like RMSHE, RSIHE , Or BPHEPL that uses mean, median, values to segment the histogram first and also other methods that apply clipping to enhance.

[Image Analyst]

OK. Good luck.

Answer 2

[soudeh]

 0 down vote favorite

Could you please help me to know what is the relation between thresholding to binary state (0,1) and noise in images? I mean that "Does thresholding introduce noise when processing images" or intensify noise?

I read somewhere that thresholding is associated with the side effect, called “salt and pepper" noise. Is there a good reference to this idea?

Thanks for your help

[Image Analyst]

Anything that is not signal is noise. So if you threshold you get a binary image (true/false, 1/0, foreground/background) that is not the image. You could call that noise but no one ever does because it's a completely different image. Salt and pepper is not a side effect of thresholding, though when you threshold you may get some small regions that you don't consider part of the objects you're interested in. You can use bwareaopen() to remove small particles after thresholding.

For what it's worth, see my attached salt and pepper noise demos.