Improved Point Cloud Guided Filtering Algorithm

This paper proposes an improved point cloud guided filtering algorithm to address the difficulties in separating and removing noise close to the model surface in point cloud denoising, as well as the problem of losing valid points during noise removal. First, the statistical filtering method is used to screen out difficult-to-smooth noise and perform initial guided filtering, thereby reducing the impact of difficult-to-smooth noise on the overall filtering effect. Then, based on the geometric features of each point in the point cloud, the weight parameters are adaptively adjusted and incorporated into the improved guided filtering algorithm for the second round of point cloud guided filtering. Finally, a smoother point cloud is obtained by adaptively adjusting the weight parameters while preserving valid points. According to experimental results, the proposed algorithm shows substantial smoothing effects on noisy point clouds. Moreover, the processed point cloud model has more prominent edge lines, and difficult-to-smooth noise can be well handled using the proposed algorithm.