Depth Image Layers Separation (DILS) algorithm of image view synthesis based on stereo vision

Abstract

A new Depth Image Layers Separation (DILS) algorithm for synthesizing inter-view images based on disparity depth map layers representation is presented. The approach is to separate the depth map into several layers identified through histogram-based clustering. Each layer is extracted using inter-view interpolation to create objects based on location and depth. DILS is a new paradigm in selecting interesting image locations based on depth, but also in producing new image representations that allow objects or parts of an image to be described without the need of segmentation and identification. The image view synthesis can reduce the configuration complexity of multi-camera arrays in 3D imagery and free-viewpoint applications. The simulation results show that depth layer separation is able to create inter-view images that may be integrated with other techniques such as occlusion handling processes. The DILS algorithm can be implemented using both simple as well as sophisticated stereo matching methods to synthesize inter-view images.