Error in image normalization

Hello Team,

I have written this code to compute the normalization of a gradient image.

if true
  % code
function [horDiffNorm, verDiffNorm] = normalization(H, V)

    sumH = 0;
    sumV = 0; 
    [rowsH, colsH, depthH] = size(H);
    [rowsV, colsV, depthV] = size(V);

    for jH = 1:1:rowsH
        for iH = 1:1:colsH
            for zH = 1:1:colsH
                if (zH ~= iH)
                    sumH = sumH + power(H(jH, zH), 2);
                end
            end
            horDiffNorm(jH, iH) = H(jH, iH)/ sqrt(double(1/(2*(colsH-1))*sumH));
         end
    end
  end
end

I performed debugging and noticed the value of SumH reaches (255) at zH = 32 and then it stuck at 255, even though the next pixels are not 0.

Can you please help?

Answer 1

[Walter Roberson]

Right after the function header put in

H = double(H) ;

[Matlab Student]

Working! Thanks Walter

Answer 2

[Image Analyst]

Why not simply use imgradient()?

[Matlab Student]

Thanks for you comment. I am new to Matlab and did not know about imgradient(). I need to perform normalization following the equation attached to this comment.

[Matlab Student]

Based on Matlab documentation, I can use this syntax: [Gmag,Gdir] = imgradient(I,method)

But the above equation is not listed under the available methods.

Am I right?

[Image Analyst]

What is that formula supposed to do? If you just want the gradient magnitude, why can't you just use imgradient()? What do you need to do AFTER that?

[Matlab Student]

The value of every pixel should be divided by the square root of the summation of neighbors to the right and left of the subjective pixel

Dnormalize(x,y) = D(x,y)/sqrt(1/2N*summation(power(D(x+i,y),2)))

N is the number of neighbors to the left and right of the pixel (x,y)

[Image Analyst]

You're going all the way from the pixel to the extreme far right edge and extreme bottom edge, instead of just in a localized scanning window like usual. Why???

[Matlab Student]

I am a very beginner to Matlab and I don't know another way to do the same. What do you mean by "a localized scanning window" ?

If you can show me another professional and efficient way to do the same, it will be appreciated.

[Matlab Student]

Would you please check this out: https://www.mathworks.com/matlabcentral/answers/422634-analog-output-signal-after-applying-dft

[Image Analyst]

What are you trying to do with this unusual algorithm? A spatial filter? A convolution? A correlation? Why did you program it up in the first place?

What is H? Is it already the gradient image, or do you want to take the gradient of H?

[Matlab Student]

I am trying to implement a method designed to measure the blockiness of distorted images. This methods was proposed in "A Blind Reference-Free Blockiness Measure" paper for Chunhua Chen and Jeffrey A. Bloom (Dialogic Media Labs) and was published in 2010.

[Matlab Student]

I have already written the code, but while testing/ debugging it still need some modifications.