Display Dependent Preprocessing of Depth Maps Based on Just Noticeable Depth Difference Modeling

摘要: This paper addresses the sensitivity of human vision to spatial depth variations in a 3-D video scene, seen on stereoscopic display, based an experimental derivation just noticeable difference (JNDD) model. The main target is exploit perception humans suppressing unnecessary details, hence reducing transmission overhead allocated maps. Based JNDD model derived, map sequences are preprocessed suppress details that not perceivable by viewers and minimize rendering artefacts arise due optical noise, where noise triggered inaccuracies estimation process. Theoretical evidences provided illustrate proposed adaptive preprocessing filter does alter visual quality or view synthesis for free-viewpoint applications. Experimental results suggest bit rate coding can be reduced up 78% maps captured with depth-range cameras 24% estimated computer algorithms, without affecting arbitrary quality.