How to filter an histogram with gaussian low pass filter ?

I have MR slice of type double, I want to filter its histogram to smooth it using Gaussian Low Pass Filter. I know that the (imgaussfilt ) filter the image with a Gaussian filter but we need the standard deviation. Could any one tell me how to fix the value of the standart deviation ? is it the (std2) of the whole image ? Otherwise, if you have you any other prepositions, I will be greatful.

[Jan]

The histogram is a 1D function, while imgaussfilt operates on 2D. What are the inputs and which output do you want?

[Image Analyst]

I'm also not sure if you want to filter the image itself, or its histogram. std2() gives the standard deviation of the entire image. stdfilt() gives the standard deviation in a moving window and thus returns an image. imgaussfilt() can filter both 2-d images and 1-d signals (such as a histogram). Why do you think you want to smooth out the histogram? Or is it the actual image , rather than the histogram of the image, that you want to smooth out (blur)?

[MMSAAH]

Thank you for your reply. I want to smooth out the histogram of the original image (which is one slice MR image ) in order to smooth the most frequent ripples of the histogram and to distinguish the major peaks and valleys in the original histogram. So that convoluate the original histogram with the Low Pass Gaussian Filter will smooth the original histogram.

[Jan]

What is the final result? A modified histogram or do you want to modify the image, such that the histogram of the modified image is smoother?

Answer 1

[Image Analyst]

Just take the histogram counts and use conv() to smooth

windowSize = 15; % Adjust to control level of smoothing.
smoothedCounts = conv(counts(:), ones(windowSize, 1)/windowSize, 'same');

[MMSAAH]

Please how to fix the windowSize ? Is it an arbitrary choice ? Because when changing value of windowSize, the curve of the histogram change

[Image Analyst]

What does "fix" mean. It IS a constant, "fixed" value, so do you have a different definition? I picked 15 arbitrarily, but like I said, you can pick different values to make the histogram smoother or less smooth (more like the original). There is not one, "right" value. It's totally up to you how much you want to smooth it.

Perhaps you'd like to use the MATLAB smoothed histogram that you get with fitdist(), if you have the Statistics and Machine Learning Toolbox.