Discontinuity-adaptive Depth Map Filtering for 3D View Generation

A depth image-based rendering (DIBR) technique is one of the rendering processes of virtual views with a color image and the corresponding depth map. The most important issue of DIBR is that the virtual view has no information at newly exposed areas, so called disocclusion. The general solution is to smooth the depth map using a Gaussian smoothing filter before 3D warping. However, the filtered depth map causes geometric distortion and the depth quality is seriously degraded. Therefore, we propose a new depth map filtering algorithm to solve the disocclusion problem while maintaining the depth quality. In order to preserve the visual quality of the virtual view, we smooth the depth map with further reduced deformation. After extracting object boundaries depending on the position of the virtual view, we apply a discontinuity-adaptive smoothing filter according to the distance of the object boundary and the amount of depth discontinuities. Finally, we obtain the depth map with higher quality compared to other methods. Experimental results showed that the disocclusion is efficiently removed and the visual quality of the virtual view is maintained.