I am trying to write a function that computes CDF of an image, but the output I get is always zero. what am I doing wrong?

function cdf = CDF(im,j)
  rows = size(im,1);
  cols = size(im,2);
  s=double(rows^cols);
  sum=double(0);
  cd
  for i=0:j
    for r=1: rows
      for co=1: cols
        pixel=im (r,co);
        if (pixel==i) 
          sum=sum+1;
        end
      end
    end
  end
  cdf = double(sum/s);

[Image Analyst]

It looks like lots of things. But let's get some information first. What is im - a gray scale image, a color image, or a binary image? What is (the very badly-named) j - its value and it's purpose? Are you just trying to count the number of pixels in the image that have value j? Is cdf the cumulative distribution function? If so, why not simply call histcounts() and cumsum()? Why are you using sum - a very important built-in function - as the name of your variable?

[dpb]

...
if (pixel==i) 
  sum=sum+1;
end

will only sum if the value of the image itself is an integer equal to the particular value of i on that pass thru the outer loop.

What is j and what is the normalization for the image? If it were scaled, the values wouldn't ever match which would cause the symptom.

You're not going to get a CDF anyway, because you're only going to return a single value.

[Reihaneh Khoshghadam]

my image is a uint8 grayscale image. no, i'm not trying to count the pixels with value of j. j is a parameter that the user has to set value to. for example if the user sets j to 20, the function is supposed to count the pixels in the picture that have the intensity of 0 plus the pixels with intensity of 1 ,...., plus the pixels with intensity of 20.

the function above does show the value of sum correctly, but when I try to get the value of cdf, which is sum/s, it becomes zero.

I believe the problem is that the value is smaller than 1, and the type of my parameter cdf or s or sum isn't correctly initialized and it floors the value to the integer 0.

Answer 1

[Steven Lord]

I think you should use histcounts or histogram. Since you want the CDF you'd specify that as the 'Normalization' in your histcounts or histogram call.

Answer 2

[Abdul Rehman]

Basically if u want to calculate cdf of image, then you have to follow three step,

1.Histogram

2.Normalized Histogram

3.CDF

Here you have a code of histogram using for loops (but you can use "Hist" function).

if true
function [hist hist_P]=histor_g(im)
hist=zeros(1,256);
[r c]=size(im);
s=r*c;
for i=1:r
  for j=1:c
      int_val=im(i,j);
      hist(int_val+1)=hist(int_val+1)+1;
      end
      end
      hist_P=hist./s; %Normalized Histogram(PDF)
     end
    end

As using this function you van get PDF(Normalized Histogram)

then you can calculate CDF.

if true
 function [c_hist]=cum_h(hp)
 [r c]=size(hp);
 c_hist=zeros(1,256);
  for j=1:c
       if(j ==1)
       c_hist(j)=hp(j);
       else
       c_hist(j)=hp(j)+c_hist(j-1);         
       end
  end
end

Hopefully you get it thanks.