An efficient Stereo matching method to reduce disparity quantization error

Efficiently utilizing the stereo Images to generate a desirable semi dense disparity map is a challenging problem. Disparity image is a projected geometric space contains more primitive information directly computed from stereo Images. Stereo matching is considered as difficult in image processing due to complexity and structure ambiguity. In this Paper, we propose a novel efficient disparity image processing to resolve the difficulties of the seed growing algorithms for small fraction of disparity space. The proposed model is computed using squared intensity and absolute intensity difference based disparity space for both low and high resolution images to achieve smoothness property. To reduce the disparity quantization error, we use multi fitting algorithm through sub pixel disparity estimation to obtain the effective and consistency in disparity mapping. The experimental results on Middlebury data’s with ground truth disparities to demonstrate that proposed method with quantitative results in order to produces high quality disparity map with less computation time and high matching accuracy along complexity Q1reduction.