How do I remove the scattered noise from my Image?

Hello,

I am trying to remove the round white bodies from my image. I think it is salt and pepper noise. But I have tried applying medfilt2 function with 3X3, 4X4 and 8X8 neighbourhood values yet I have not been able to remove these white spot-like circles. Could anyone please help me how to remove this noise?

Thanks in advance!

Answer 1

[KALYAN ACHARJYA]

You can do that in multiples ways- Use morphological operations (Documentation here)

im=imread('image3.jpeg');
im2=im2bw(im);
figure, imshow(im2);
im3=bwareaopen(im2,60);
%..................^.. Change this value as per your requirements
figure, imshow(im3);

You can also look here bwareafilt

[Guru Charan Das Mudigonda]

Hello,

Thank you so much! That solved my problem. But I have some doubts. I used bwmorph to perform pixel wise morphological operations before posting this question and yet I have not got the results. I have used opening, cleaning and removing operations but did not get the desired result. Why did these operations not work? And also, why did the median filter not work? I have seen in the MATLAB medfilt2 documentation that this filter is used to remove salt and pepper noise. Why did this not work then?

Thanks in Advance!

[KALYAN ACHARJYA]

openng and closing (Based on Eroison and Dilation) will apply on all white pixels, it doesnot matter about the objects sizes, therefore its effiect on all whilte objects.

If you apply the logical cleaning operations in proper flow, certainly it works.

Regards,

[Image Analyst]

You accepted the answer but from your last comment it sounds like it didn't work. So did you figure it out?

Median filter might not have worked if you used the wrong window size, and if you used a bigger one it mgiht blur the object you want to keep.

If you want to get the cleaned gray scale image you'll have to mask it

cleanedGrayScaleImage = im; % Initialize
cleanedGrayScaleImage(~im3) = 0; % Make/erase noise from original image.

Attached are my salt and pepper filters but I doubt they'll work for your images because I don't think your noise is like that.

[Guru Charan Das Mudigonda]

Hello,

Thank you for your answer. Yeah that did work for me, using bwareaopen function. But my doubts completely were not clarified. If you could, could you just explain me why the median filter won't work for this type of noise, it would be really helpful. Also, like Kalyan Acharya has said, if logical cleaning operations could be applied to obtain the similar result after de-noising, how could I do it in a 'proper flow'?

Thanks in advance!

[KALYAN ACHARJYA]

- just explain me why the median filter won't work for this type of noise-

Suppose you have gray image

Pixels: 20, 34 ,40,45, 100,110,120,200,255

Center Pixel replace by median: 100

As from the image, the noises & its adjacent boundary also having similar pixel values. It doenot effect iy you choose the smaller window, like 3x3 (most noises-white small dots are much larger than 3x3 pixels). If you choose the window size more than 3x3 (say 15x15), I expect it will works. But its effects on non noise parts also.

see the following example:

im=imread('image3.jpeg');
figure,imshow(im);
im2=medfilt2(im,[15 15]);
figure,imshow(im2);

Lets wait for better clarification from @Image Analyst sir

[Guru Charan Das Mudigonda]

Hello,

Thank you for your response. I understand the logic now. I will try the 15X15 neighbourhood. But like you said, I think even the important part of my image that must not be filtered would be blurred with this neighbourhood. Of course, if possible, I would like to know how to use cleaning and removal morphological operations in correct flow to obtain the same result.

Thanks in advance!