Distance transform-based depth map upsampling

초록

This paper presents distance transform-based bilateral upsampling that converts a low-resolution depth map into its high-resolution one. A bilateral upsampling has a disadvantage that the sharpness of edge boundary is not maintained. To overcome this, the distance transform that is defined as the distance of each non-edge pixel from the nearest edge pixel is incorporated into the bilateral upsampling. Combining the distance transform value and a spatial weighting function produces a high resolution depth map with edge sharpness preserved. For performance evaluation, we implemented six conventional depth map upsampling methods. Experimental results performed on ten depth maps show that our method outperforms the conventional methods on average by 1.0 db in terms of the PSNR (peak signal-to-noise ratio). © 2014 ISSN 1881-803X.

키워드