Hi Yin , generally speaking, you can use lots methods to extract texture features. You can collect a huge number of descriptors such as Tamura descriptors, statistical metrics (mean, variance, skewness and kurtosis), fourier transform, local keypoints descriptors (SIFT and SURF mostly).

You can find out which among the aforementioned descriptors is more useful depending on your objective and topic. If you give some details I could give some suggestions. Bye.

hi, Bruno, thanks your answer. when i do the edge-preserving image filtering in a image, i need a method to texture extraction. do you have some suggestions?

If your objective is to analyse the smoothness or the roughness of the image after the filtering, you could try to account the level of roughness starting from easy and off-the-shelf descriptors such as mean, variance, skewness and kurtosis, in order the statistical descriptors of first to fourth order.

You can find the pre-built code in Matlab coding (or you can easily find some code on mathworks website). My suggestion is to make some attempts with statistical descriptors for now by using the nlfilter of matlab (nlfilter is used to apply some nonlinear filter function) to account the variations of the texture into the image.

Then you could make some other attempts by using Tamura texture descriptors (more complex from a computational viewpoint).

Let's try and Let me hear from you if this approach is useful for your purpose.

Contact me if something I wrote is not clear.

Bye,

Alessandro

thankyou, I just saw your reply. i will try . if this approach is useful , i will let you know.

bye,

yin