pixel bin relation imhist

Hello,

I'm using imhist, and I would like to assign a default value to each pixel whose intensity appears less than a fixed number of times in the image. In other words, I would like a function whose input is an image and the output is its histogram and the concerned pixels (their location) for each bin of the histogram. I could write a script doing that, but I'm looking for a direct method, as optimized as possible.

Thanks for your help

Sylvain

[Adam]

Well, I'm pretty sure there isn't a single inbuilt function that will do it all for you so you'll have to write some code. It depends what you consider a 'direct' method vs whatever you think the alternative is.

Answer 1

[Jan]

The question is not clear. What are " concerned pixels (their location)" exactly?

I assume you want to start with:

[counts, binLocations] = imhist(I);   % Or imhist(I, n) with a suitng n?

And now you want set all binLocations with counts<limit to a certain value? How would the script you have mentioned look like?

[Sylvain Hauser]

I was looking for a Matlab built-in function or a very short script which gives me, for each bin of the histogram, the positions (x,y), of the pixels belonging to this bin. I don't want to modify the histogram, but the image itself. For now, I wrote a script which is not ideal, since it has a for loop in order to access to each bin of the histogram. But apparently, there is not necessarily a better way to do.

[Adam]

Well, you haven't shown your script so who knows if there is a better way or not until you do?!

[Sylvain Hauser]

Sorry, here is my script, f being the grayscale image and F a factor to determine the threshold:

function [f] = imhist2(f,F)
bl=linspace(0,1,256);
bl1=zeros(size(bl));
bl1(2:end)=(bl(1:end-1)+bl(2:end))/2;
bl2=ones(size(bl));
bl2(1:end-1)=(bl(1:end-1)+bl(2:end))/2;

th=F*max(imhist(f));

cn=zeros(size(bl));
for i=1:length(bl)
    ind=find(f>bl1(i)&f<bl2(i));
    cn(i)=length(ind);
    if cn(i)<th
        f(ind)=mean2(f);
    end
end

end

Answer 2

[Jan]

Try this:

Edge    = linspace(0,1,256);
[N, Edge, Bin] = histcounts(f, Edge);
V       = (N < (F * max(N)));
Mask    = V(Bin);
f(Mask) = mean(f(:));

histcounts determines the frequency of values. I've used the simpler linspace(0,1,256) here, because the intention of your bl1 and bl2 is not clear to me. You can adjust the wanted edges as you want. Then a mask is created, which is TRUE for all values, which appear less than F times the most frequent value.

Answer 3

[Image Analyst]

You can compute the color frequency image http://www.mathworks.com/matlabcentral/fileexchange/28164-color-frequency-image

Then simply threshold it and assign the new value you want.

pixelsToReplace = cfImage < someThresholdValue;
cfImage (pixelsToReplace) = desiredValue;