For the co-occurence (2nd order) texture calculations, how is correlation handled when the window and shifted window have all the same binned values (i.e. all 1's)? Given the formula for correlation, you cannot calculate correlation because there is 0 variance. It appears from testing in ENVI that a value of exactly 1.0 is returned for this case (although I could be wrong). Other software I've used will return a nodata value. If indeed this is a choice by ENVI software developers to insert a value of 1, then it may be prudent to clarify in the documentation. Thanks.