Disparity estimation can be used to eliminate redundancy among different views in multi-view video compression to obtain high compression efficiency. However, the problem of high computational complexity in disparity estimation, which limits real-time applications of multi-view systems, needs to be solved. In this paper, a novel fast multi-view disparity estimation algorithm based on Hadamard similarity coefficient for multi-view video coding is proposed by using prediction of initial search point, selection of reference view, determination of the best disparity vector, and strategies of search stop. Experimental results show that the proposed algorithm can significantly reduce the computational complexity in multi-view disparity estimation.