Question has an ambiguous form. The texture may be MORPHOLOGICAL, than it is answered above.. However if the crystallography (lattice orientation) is implied (colored mapping of orientations by EBSD is a usual case) everything changes!   No answer for the latter as yet since the question should be cleared first. 

Read a book on image analysis.

I'm not sure if I understand you question, but for detecting textures in grey-scale images it's common to use a feature called LBP (Local Binary Pattern) and it exists some generalized versions of it that allows to improve the performance through color information like this one: http://liris.cnrs.fr/Documents/Liris-5190.pdf

Have a look at Law's texture energy filters.

As mentioned before, your question is first to define a "texture feature" or more properly, extract a texture measurement from your data. There are numerous approaches to texture:

• Gabor Filters,
• Co-occurrence,
• Trace Transform,
• frequency filters,
• Laws Masks,
• Wavelets,
• LBPs,
• Order Pyramids, etc. 

Once you have identified the approach or measurements, then you can proceed to extract them and do a feature selection. Take a look at this document as a good tutorial, it is for 3D textures but all is applicable to 2D.

http://eprints.dcs.warwick.ac.uk/747/