Performance improvement of segmentation-based depth representation in 3D imagery by region merging

The feasible implementation of immersive 3D video systems entails the need for a substantial reduction in the amount of image information necessary for representation. Multiview image rendering algorithms based on depth data have radically reduced the number of images required to reconstruct a 3D scene. Nonetheless, the compression of depth maps still poses several challenges due to the particular nature and characteristics of the data. To this end, this paper outlines a depth representation technique, developed in our earlier work, that exploits the correlation intrinsically present between color intensity and depth images capturing a natural scene. In this technique, a segmentation-based algorithm that is backwards compatible with conventional video coding systems is implemented. The effectiveness of our previous technique is enhanced in this contribution by a region merging process on the segmented regions, which results in a decrease in the amount of information necessary for transmission or storage of multiview image data by a factor of 20.5 with respect to the reference H.264/AVC coding methodology. This is furthermore achieved whilst maintaining a 3D image reconstruction and viewing quality which is quasi identical to the referenced approach.