Hii everyone, how can I calculate histogram based first order statistical features from an image (such as average grey intensity, skewness, kurtosis, entropy, varience, std. deviation, etc.). Also how to calculate component based intensity features.

Hii everyone, how can I calculate histogram based first order statistical features from an image (such as average grey intensity, skewness, kurtosis, entropy, varience, std. deviation, etc.). Also how to calculate component based intensity features . What is the difference between the two sets.

Answer 1

[Image Analyst]

You can use mean(), mean2(), var(), var2(), std(), std2(), skewness(), entropy(), etc. For more general image moments, see my attached demo.

For feature based measurements, use regionprops() as shown in my Image Segmentation Tutorial: http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862

[surbhi pareek]

Hii image analyst thanx for the help... Could you please elaborate on the difference between first order histogram based features and component based intensity features. I am attaching the text from the reference paper where I got to know about this. Thanx in advance. {.3. Component based intensity features Component based intensity features are computed from the pixel intensity values of connected components in the image. The features are almost similar to the histogram based first order statistical features. The difference between two is in size of image. Histogram based first order statistical features were derived from 256 256 size image, while image size in the time of component based intensity features extraction depends on the size of con- nected components. In this research, maximum intensity, mini- mum intensity, average intensity, median intensity, standard deviation and variance of intensity values are computed from com- ponents. The extracted features are defined as follows:

where g is the number of distinct grey level in the defined con- nected component. p(g) is the histogram of the pixel intensity of connected component and l is the possible intensity level of the connected component. Mean intensity is the average pixels value, which is determines the brightness or darkness of the defined con- nected component. Maximum and minimum intensity define the maximum and minimum pixel intensity value of the connected component respectively. Variance determines the dispersion of grey level pixels from the average within the connected component and standard deviation determines mean square deviation of grey level value from its mean value. If the intensity values are arranged in ascending order, then middle value is defined as median intensity value

[Image Analyst]

Let's say you wanted to know the penetration of some dye into some objects soaking in a solution. Let's say that the average dye penetration in from the outside edge of the objects as a profile plot is something you're interested in. There's no way you can get that strictly from the histogram or moments. Let's say you wanted to know how many holes are in your blobs. You can get that from the Euler number from regionprops but there is no way you can get that from histograms. Let's say you wanted to know the Solidity or aspect ratio or any number of other things. You can't get them from the histogram or from moments.