Multiview extension of HEVC using reduced resolution disparity information for inter-view prediction tools

This paper investigates the effect of reducing the resolution of estimated depth maps that are used for inter-view motion and residual prediction in a multiview extension of HEVC. The investigated HEVC extension includes disparity compensated prediction similar to the MVC extension of H.264/AVC. For further increasing the coding efficiency, the obtained disparity vectors are used to estimate depth maps, which are employed by two additional tools: The inter-view motion prediction uses the estimated data for predicting motion vectors of dependent views from the independent view while the inter-view residual prediction uses the depth data for predicting the residual data of dependent views. Simulations showed that, on average, bit rate savings of about 10% for the dependent views are achieved by enabling the additional coding tools. By reducing the resolution of the estimated depth maps, the decoding complexity overhead associated with the inter-view motion and residual prediction can be significantly decreased from about 167% to 128% at the expense of a minor bit rate increase of about 0.5%.