In this paper we present a new theory of possibility for correspondence search. The proposed theory is based on fuzzy modeling of stereoscopic constraints. It defines a possibility of matching and a possibility of unmatching for each pair of pixels from the two stereo images. We estimate the initial disparities using a confidence degree calculated using the proposed possibilities. An adaptive support window, whose weights are computed using initial disparities, is used in a SAD aggregation function to calculate final disparity map. Experimental results and comparison with other algorithms are presented to demonstrate the performance of our approach.