We have improved our Concurrent Stereo Matching (CSM) algorithm, which abandons the search for ‘best’ matches and determine matches that lie within admissible ranges using a noise model. We estimate photometric deviations between corresponding regions of stereo pairs with photometric transformations and mismatched or occluded regions. We allow for global, disparity dependent contrast and offset (gain and dark noise) distortions as well as multiple outliers. Noise is estimated for each pixel at each disparity level and the CSM framework applied. Outliers are eliminated with a statistical model and likely matching volumes identified. Then, starting in the foreground, the volumes are explored to select mutually consistent optical surfaces. Finally, local, not global, surface continuity and visibility constraints are applied to generate a disparity map. This approach compares well with other matching algorithms: the more realistic matching model allows for signal contrast and offset variations over the whole image.