问题
Assume that I have a grayscale image. Consider a circular neighborhood window around each pixel. I need to get color histogram of circular neighbourhoods around each pixel.
How can I efficiently implement circular neighborhood operations for this problem in MatLab?
回答1:
I don't want to give you everything, but I think this should help you out a lot.
Well you can make a circle of ones doing something like
h = fspecial('disk',rad);
h = h>0;
Then you can put that anywhere in a larger matrix doing something like
h2 = zeros(N,M);
h2(c_offset-rad:c_offset+rad,r_offset-rad:r_offset+rad) = h;
Now you have a matrix the same size (col/row size) as your image. You can use this as a reference table for getting data from a matrix, much in the same way you can return only the values above 0.5 by saying
r = rand(10);
d = r(r>0.5);
EDIT:
You'll also need to play around with the data types in some places to make MATLAB happy. For example, h2 will need to be a logical to use it as a reference table for another matrix. And hist won't work without proper types either.
来源：https://stackoverflow.com/questions/13233949/circular-neighborhood-operations-matlab-color-histogram