I am not sure if there are any standard algorithms to perform this task. But you can try a Bayesian approach, where you have to maximize the posterior probability, p_o(L(x)=a|I_t(x),I_s(x),~L(x)) with respect to the texture image, I_t(x), the stereo image, I_s(x) and a smoothness term~L(x) (which could be a 3x3 neighborhood pixel labels at image position x). You can try gaussian distributions for the likelihood and prior terms respectively.
If you want to segment image using texture, i suggest you to try LBP(Local Binary Pattern) which introduced for pattern analysis in computer vision.Also,It has many variations and developments.
Depth information from stereo images can be used for classification tasks.This Method is very new and recently used by one author for Indoor Segmentation.