No-reference quality assessment method of color point cloud based on the combination of 3D and 2D information
As a carrier of 3D visual information,point clouds inevitably exhibit distortions that are introduced during their acquisition,transmission,and reconstruction,leading to a reduction in their visual quality.Therefore,it is necessary to effectively measure the quality of distorted point clouds.This paper proposes a no-reference quality assessment method of color point cloud based on the combination of 3D and 2D information,with which the geometric information and color texture information of point clouds in 3D and 2D spaces are extracted,respectively.To account for the combined distortion of geometry and color texture in point clouds,a tensor decomposition-based approach is applied to jointly analyze and extract features from the combination of these two projection maps.Considering the multi-directional and multi-scale perceptual characteristics of the human eye,a curvelet transform is used to extract features from the color texture projection map.Finally,a random forest pooling method is employed to predict the quality of color point clouds.Experimental results on three databases demonstrate that the proposed method outperforms some of the existing point cloud quality assessment methods,in that it shows better correlation between the predicted results and subjective quality assessment.
color point cloudquality assessmentcombination of 3D and 2Dperceptual feature
NETL
NSTL
万方数据
