Parameter optimization for point clouds denoising based on no-reference quality assessment

Almost all point clouds denoising methods contain various parameters, which need to be set carefully to acquire desired results. In this paper, we introduce an evolutionary optimization algorithm based framework to obtain the parameter configuration of point clouds denoising methods automatically. New no-reference quality assessment metrics are proposed as objective functions to quantitatively evaluate the point clouds during the optimization process. The proposed metrics infer the quality of point clouds in terms of both smoothness and density. The ideas of manifold dimension and holes detection are combined to get the smoothness evaluation results. Simplified local outlier factor is further exploited for the density evaluation. Using public dataset and real-world scanned data, experimental results prove that the automatic tuning parameters provide a significant boost in performance compared with the manual tuning parameters. Furthermore, the results acquired by the proposed metrics achieve better or equivalent performance than the state-of-the-art metrics.