what is MAD of an image? on which basis it calculated the difference between images?

Can anybody please explain MAD of an image caculated on which basis? how it calculate mad on pixels?

Answer 1

[Bjorn Gustavsson]

I have a statistics toolbox function MAD that calculates either the mean or median absolute deviation. That is:

MEAN(ABS(X-MEAN(X))

or

MEDIAN(ABS(X-MEAN(X))

Can't be bothered to read more of the help for MAD, so cant tell you how it handles matrices, but if it works on columns and you want the MAD of the entire 2-D image use the ":" operator:

MAD(X(:))

HTH,

[karthikeyan krishnamoorthi]

Thank you so much sir.

Answer 2

[Image Analyst]

I have a demo for computing the MAD image. I use it to find outliers and replace them, thus repairing a noisy image. I'll upload it later this morning once I get to my other computer. I don't use the Stats Toolbox - I just use the median function and nlfilter() from the Image Processing Toolbox to filter each pixel based on a local neighborhood. Basically MAD is a method for finding outliers, so I found the outliers (noise) and replaced them with good pixels (from the median filtered image).

[Image Analyst]

See attached demo.

[Bjorn Gustavsson]

Well, it is also an alternative to standard deviation for measuring spread of data with long tails in the distribution around the mean/centre.

[karthikeyan krishnamoorthi]

thank you very much sir