A novel region-based progressive stereo matching algorithm is presented. It combines the strengthes of previous region-based and progressive approaches. The progressive framework avoids the time consuming global optimization, while the inherent problem, the sensitivity to early wrong decisions, is significantly alleviated via the region-based representation. A growing-like process matches the regions progressively using a global best-first strategy based on a cost function integrating disparity smoothness and visibility constraint. The performance on standard evaluation platform using various real images shows that the algorithm is among the state-of-the-art both in accuracy and efficiency.