An adaptive motion estimation algorithm based on mutual information for depth information estimation

Left and right eye's views could be derived from the monoscopic image and corresponding depth map. Presenting these two slightly different views of every scene to the individual eyes with an appropriate disparity, a correct 3D perception can be realized within the mind. Taking the advantage of motion vector which reflects the movement between neighboring frames, the depth map could be reconstructed. The motion vector calculated by traditional motion estimation algorithm is not appropriate for depth information estimation and the computational requirement is huge. So we introduce the parameter of mutual information (MI). The block of motion estimation is classified into one of the two different types corresponding to background and active moving block depending on MI. The proposed adaptive method exploits the mutual information and adjusts the size of search area and block size. Simulation results show that the proposed algorithm can save the coding time by 10% compared with the full search algorithm and the quality of the depth map is improved significantly.