Multi-Level View Synthesis (MLVS) Based on Depth Image Layer Separation (DILS) Algorithm for Multi-Camera View System

A novel Multi-Level View Synthesis (MLVS) approach for 3D vision and free-viewpoint video applications, such as light field imaging, is presented. MLVS exploits the advantages of Depth Image Layer Separation (DILS), a new inter-view interpolation algorithm, by extending stereo to multiple camera configurations. The technique finds the pixel correspondences and synthesis through two levels of matching and synthesis process. The main aim of MLVS is to create a multi-camera view system through a reduced number of actual image acquisition cameras, whilst maintaining the quality of the virtual view synthesis images. The proposed technique is shown to offer improved performance and provide additional views with fewer cameras compared to conventional high volume camera configurations for free-viewpoint video acquisition. Thus, substantial cost savings can ensue in processing, calibration, bandwidth and storage requirements.