Subjective and objective quality evaluation of 3D point cloud denoising algorithms

Abstract

The increasing availability of point cloud data in recent years is demanding for high performance denoising methods and compression schemes. When point cloud data is directly obtained from depth sensors or extracted from images acquired from different viewpoints, imprecisions on the depth acquisition or in the 3D reconstruction techniques result in noisy point clouds which may include a significant number of outliers. Moreover, the quality assessment of point clouds is a challenging problem since this 3D representation format is unstructured and it is typically not directly visualized. In this paper, selected objective quality metrics are evaluated regarding their correlation with human quality assessment and thus human perception. As far as the authors know, this is the first paper performing the subjective assessment of point cloud denoising algorithms and the evaluation of most used point cloud objective quality metrics. Experimental results show that graph-based denoising algorithms can improve significantly the point cloud quality data and that objective metrics that model the underlying point cloud surface can correlate better with human perception.