How to identify the type of the Histogram?

Hi All,

I have medical images,I've plotted it's histograms,now I need to identify the type of these histograms(e.g. One may be like Gaussian distribution, Normal distribution.etce...),is there a builtin Matlab function or any code does that?thank All in advance for time and support.

[Walter Roberson]

Duplicated by newer http://www.mathworks.co.uk/matlabcentral/answers/112617-how-to-identify-the-type-of-the-histogram-distribution which should be merged into this.

Answer 1

[Mischa Kim]

Check out this tool on File Exchange. Alternatively, ktest2 might be an option.

[hamed abdulaziz]

Thank you Mr.Mischa Kim for quick answer,in fact my goal of this qestion is: I have two algorithms for image segmentation,so if the distribution was Gaussian type I will use the first algorithms else use the second algorithm,please could you guide me to do this job.thank in advance for time and support

[hamed abdulaziz]

I am waiting your comment,please guide me

[Walter Roberson]

As Wayne King answered in your duplicate of the question, the solution for what you ask here is kstest as that tests whether the data comes from a normal (== Gaussian) distribution.

[hamed abdulaziz]

I tried this code:

test_cdf = makedist('normal','mu',10,'sigma',1);

Image1 = imread('d:\200.bmp');

[x1,counts1] = imhist(Image1(:,:,1));

stem(counts1,x1);

[h] = kstest(counts1,test_cdf);

my question is : should I get h=1 if the data seta are fit?

[hamed abdulaziz]

please I am waiting your answeres...

[Walter Roberson]

h = kstest(counts1, 'CDF', test_cdf );

would return 1 if there is a 95% or greater likelyhood that the data came from the given cdf.

Note: When you wait 3 1/2 days to reply, it is not fair to expect the volunteers to reply within 1 hour. We cannot be checking every hour for days on end. We are still debating whether Wayne King or Image Analyst ever sleep, but we have it on good authority that they do stop to eat.

[hamed abdulaziz]

Thank you, and I am sorry for bothering Mathworks community,

Please do you mean there is a probability of the similarity equals to 95% ?

[Walter Roberson]

If h returns as 1, then the probability of a false match is less than 5%, so the probability of a correct match is at least 95% (not exactly 95%)

[hamed abdulaziz]

Ok, but I tried another binary image (bit depth = 1 bit) bitmap type, and also h returns as 1

[hamed abdulaziz]

I was suppose I will get a different value of h (i.e h will be equal zero)

[Walter Roberson]

imhist returns [counts, x] not [x, counts]

For a binary image (datatype logical) you would only have 2 bins, so you would be working out the probability that the two counts are consistent with a Gaussian distribution. With only two samples, the probability calculations are not statistically meaningful.

[hamed abdulaziz]

Thank very much Dr.Walter Roberson,for clarification,and I promise you if I asked another question I'll be patient for waiting reply . Regards