Histogram Equalisation creating pixelated images

Question

jlouise2022 2022-7-5

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1754185-histogram-equalisation-creating-pixelated-images

评论： jlouise2022 2022-7-6

Hi there,

I am using histeq to transform a set of images. However, while some of the images are transformed fine, a fair few are looking too pixelated. I will input an example below. First, here is my code:

filenames = dir(fullfile(Stimuli, '*.jpg')); 
for k = 1:numel(filenames)
  images   = filenames(k).name;
  equalHist     = imread(images);
  gsImg      = histeq(equalHist, 255);
  [~,name,~] = fileparts(filename);
  gsFilename = sprintf('%s_histeq.jpg', name);
  imwrite(gsImg,gsFilename);
end

This is how some of the images are looking. E.g. edges are pixelated.

I have tried adjusting the intensity availibility , e.g. 255, 127, etc, but the problem persists.

Any ideas welcomed!

Thanks

2 个评论
显示无隐藏无

DGM 2022-7-5

在 MATLAB Online 中打开

shallowgrad.jpg

Two things:

your images are integer data
your images are lossy-compressed JPG

What happens when you take a roughly-quantized shallow gradient and stretch its range?

A = imread('shallowgrad.jpg'); % a shallow JPG gradient

A = rgb2gray(A);

imshow(A)

B = imadjust(A); % stretch the range

imshow(B)

IA is right about not using histeq(), but the problem is more fundamental. Images have limited information. They look grossly quantized because they are. Regardless of the method, extreme contrast changes will emphasize the deficiencies of the source image.

jlouise2022 2022-7-6

Thanks for the explanation!

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Image Analyst 2022-7-5

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1754185-histogram-equalisation-creating-pixelated-images#answer_1000965

Yes, you've learned that histogram equalized images look crappy. It's not surprising. So just don't do it. There is no need to. There are ways to get it to look less posterized (by adding a half gray level of noise) but they will still look bad so don't bother.