Stereo matching is used in several applications. Due to occluded and mismatched regions, it is still considered as a difficult problem. An efficient cost for matching and an adaptive support weight technique for cost aggregation to obtain raw disparity is proposed in this paper. Finally, the raw disparity is refined to improvise the stereo matching performance in occluded and mismatched areas. Initially, an effective matching cost technique based on census transform is applied. This matching cost can withstand any radiometric variations. Next, an optimized Adaptive support weights framework for cost aggregation is used. In this framework, a joint bilateral filter and a guided filter are combined to form collaborative cost volume filtering. Finally, a novel multistep refinement technique, including outlier detection, interpolation and inpainting is proposed. The experimental outcomes of the proposed method outperforms currently existing techniques on Middlebury datasets.