Reason for Histogram Equalisation for having "stacked" effects?

I read in an image, run histeq on the it and store it as Img2. Then I run histeq again on Img2 and store it as Img3.

It turns out that when I compare the histograms of Img2 and Img3, the histogram of Img3 looks more equalised and uniform. Both Img2 and Img3, however, look almost identical with naked eyes.

This is a screenshot of their histograms. The one on the left is the histogram for Img2, while the right one is for Img3:

So it looks like the histogram equalisation can be "stacked" and by running many times can make the image even more uniform? Is there a reason for this?

Answer 1

[Image Analyst]

What was your second input argument to histeq()?

[Charles]

I did not provide a second input argument into histeq(). I only provided the image data from imread() into histeq() as its first argument and that's all.

[Image Analyst]

Why don't you try providing a flat histogram as a target and see how it does?

[Charles]

oh wait a minute, I just realise that I put 255 as my second argument. Would this matter?

[Image Analyst]

Let's take a step back and figure out why you want to do histogram equalization anyway. Almost always it's unnecessary to the subsequent processing and analysis, and produces a lousy, unnatural looking image.