A fast layer-based multiview image coding algorithm

In [1,2] we presented a multiview image coding scheme. The approach is based on extracting depth layers from multiview images. Each layer is related to an object in the scene and is highly redundant. We exploit this redundancy by using a properly disparity compensated Wavelet Transform, followed by quantisation and entropy coding of the transform coefficients. In addition, to reconstruct the data we encode a 2D contour associated with each layer. In this paper we describe extensions to this coding scheme aimed at reducing the complexity of the algorithm, while retaining its rate-distortion (RD) performance. In particular, we propose a novel low-complexity scheme to encode the layer contours. Simulation results show that the approach remains competitive with the original method. In addition we show that the reduced complexity algorithm still outperforms H.264/AVC.