Motion and Disparity Estimation with Self Adapted Evolutionary Strategy in 3D Video Coding

Real world information, obtained by humans is three dimensional (3-D). In experimental user-trials, subjective assessments have clearly demonstrated the increased impact of 3-D pictures compared to conventional flat-picture techniques. It is reasonable, therefore, that we humans want an imaging system that produces pictures that are as natural and real as things we see and experience every day. Three-dimensional imaging and hence, 3-D television (3DTV) are very promising approaches expected to satisfy these desires. Integral imaging, which can capture true 3D color images with only one camera, has been seen as the right technology to offer stress-free viewing to audiences of more than one person. In this paper, we propose a novel approach to use evolutionary strategy (ES) for joint motion and disparity estimation to compress 3D integral video sequences. We propose to decompose the integral video sequence down to viewpoint video sequences and jointly exploit motion and disparity redundancies to maximize the compression using a self adapted ES. A half pixel refinement algorithm is then applied by interpolating macro blocks in the previous frame to further improve the video quality. Experimental results demonstrate that the proposed adaptable ES with half pixel joint motion and disparity estimation can up to 1.5 dB objective quality gain without any additional computational cost over our previous algorithm. Furthermore, the proposed technique get similar objective quality compared to the full search algorithm by reducing the computational cost up to 90%.