关键词: RGB-D, 深度修复, 深度置信度, 分层联合双边滤波

Abstract:

The depth map captured by RGB-D sensor usually has low resolution and lack of depth value, but three-dimensional multi-target detection and recognition need to obtain high-precision and high-resolution depth maps. This paper proposes a hierarchical joint bilateral filtering depth map restoration algorithm based on depth confidence. The depth degradation model based on the problem of the depth information capture device is proposed. The depth confidence measurement is used to classify the depth pixels, and the filter window weight value is determined according to the depth confidence. The hierarchical joint bilateral filtering is used to complete the restoration of hole region. The qualitative comparison and quantitative result analysis based on Middlebury standard database and the proposed database show that the edge of depth map is more clear and reasonable, and there is no edge blur and texture artifacts basically. Therefore, the experimental results show that the proposed algorithm can improve the accuracy of depth map restoration effectively.

Key words: RGB-D, depth restoration, depth confidence, hierarchical joint bilateral filtering