# What information does the abs(normalized_image - normalized_quantised_image) carry?

Given a grayscale image X2

And its (under)quantised version with 3 bits depth X3

What information does the abs(X2_normalized - X3_normalized) carry?
As an example the following image X5 generated with the code below

X2norm= X2-min(X2(:)); X2norm = X2norm / max(X2norm(:));
X3norm= X3-min(X3(:)); X3norm = X3norm / max(X3norm(:));
X5 = abs(X2norm - X3norm);
image5=figure;imagesc(X5);title('image 5');
saveas(image5,'image5.jpg');


• I would suggest that you exchange the word "quantified" for the word "quantised". – A_A May 8 '18 at 7:09

Assuming your image is a $K$ bits depth image, and the quantized version is a $3$ bits version, pixels of the difference image, represent $(K-3)$ least significant bits of pixels of the original image. It means it completely is similar to original image in darker parts and completely masks the brighter parts.