3D point cloud quality assessment method using Mahalanobis distance
Abdelouahed Laazoufi, Mohammed El Hassouni, Hocine Cherifi · 2022
In this work, we suggest a Reduced-reference (RR) method assess the visual quality of a deformed point cloud with a reference point cloud of ideal quality. To do so, we extract geometrical and perceptual attributes of both reference and distorted PC. Estimates of their statistical properties (Entropy, Mean, Standard Deviation, Median, Kurtosis, and Skewness) form a features vector for each PC. The perceptual metric between two point clouds is computed using the Mahalanobis distance between their feature vectors. Finally, the random forest regressor is employed to estimate the quality score prediction. To validate our method, a set of experiments are conducted on an open subjective colored point cloud dataset (SJTU-PCQA). The results show that the suggested quality assessment method surpasses some contending methods in regards to correlation with average opinion scores.