A search range prediction for fast stereo matching in 3-D video

In this paper, we generate the valid search range information for prediction of search range. We compare a previous frame with a current one for acquiring disparity levels of pick values and the sum of object pixel values in each frame. The result obtained from comparing two frames is used to estimate the search range for the next frame. Also, to reduce the processing time, we use sub-sampled images and reuse the search range to realize the real-time system. As a result of our experiment, the processing time of our proposed algorithm is reduced by about 70% compared with the one using the full search range, with achieving similar PSNR.