How to calculate "CONTRAST PER PIXEL(CPP)" of an image

To measure the quality of an image i'm using Contrast Per pixel(CPP).

It is defined as the average intensity difference between a pixel and its adjacent pixel.

I'm not used to with this metric. If anybody is well known with this metric, please share it.

Thanks

Answer 1

[Image Analyst]

Something like this perhaps (untested)

kernel = [-1, -1, -1, -1, 8, -1, -1, -1]/8;
diffImage = conv2(double(grayImage), kernel, 'same');
cpp = mean2(diffImage);

[Tuhin Sheikh]

Could you please explain the program.

Hello sir,

i'm doing a comparative study among Global Histogram Equalization(GHE),

Contrast Limited Adaptive Histogram Equalization(CLAHE) and Brightness

Preserving Bi-Histogram Equalization(BBHE).

The output i get clearly shows that BBHE is the best among these three.

But i cannot conclude based on visualization.

I'm confuse which metric i should use with proper justification. Could you

please suggest me.

[Image Analyst]

Conv2() is a sliding filter. I made a 3 by 3 window that basically takes the (center - upper left) + (center - top left) + (center - upper right) + (center - left) + (center - right) + (center - lower left) + (center - bottom left) + (center - lower right). That is the sum of all the differences between the center and the 8 surrounding pixels. Then I divide by 8 to get the average difference between a pixel and it's neighbors. The conv2() function just does that operation on every single pixel in the image as it slides that window along pixel-by-pixel and carries out those operations. Then we have an image where each pixel is the average difference between the original pixel at that location and the 8 neighbors. The mean2 simply gets the mean over all the pixels.

You would have to have a metric that models human perception. I'm not aware of any, but you might check here: http://en.wikipedia.org/wiki/Video_quality

[Tuhin Sheikh]

Thanks for the explanation and the suggestions sir

Answer 2

[shree san]

Dear sir/Madam, Thank you so much for the above code. It helped me to calculate the contrast per pixel. But my output gives negative values. Is that right and how I can say the best contrast enhanced image from the output?

output I got is -58.3770 original image

 -499.6260  contrast enhanced image

Looking for a favourable answer. Thank you

[Image Analyst]

Yes, it is possible to get a negative Laplacian for some pixels. I don't know how you got your numbers. To get a nice contrast adjustment, try imadjust().

[shree san]

Thank you so much