Disparity estimation is used to eliminate redundancy among different views in multi-view video coding in order to obtain high compression efficiency. In this paper

characteristics of image block are extracted by using fast Walsh transformation

and a new method for reference view selection in multi-view disparity estimation is presented

based on the characteristic vector of image block and its similarity measurement. Experimental resuhs show that the proposed method can reduce the computational complexity in multi-view video coding with almost the same quality of reconstructed images.