.png){kind=icon}
.png){kind=icon}
Return
Enhanced point cloud denoising and smoothing via improved statistical denoising algorithm and global noise probability-based mean shift algorithm
.png){kind=icon}
PRE
AI
L
Y
G
X
DOI：10.1117/1.OE.65.4.048102.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Point cloud noise detection and removal are crucial for accurate 3D reconstruction. Traditional statistical denoising algorithms struggle to effectively eliminate cluster noise, and mean shift smoothing often leads to oversmoothing and vertex drift. We propose a point cloud processing method, combining a principal component analysis-improved statistical denoising algorithm to remove outliers and clustered noise, and a mean shift algorithm based on global noise probability density to regulate the drift process of noise points. Smoothing experiments show the average reconstruction errors of the Bunny and Horse models are 0.0017 and 0.0037 mm, respectively, which are significantly lower than those of Laplace smoothing and bilateral filtering algorithms. Subsequent ablation studies confirm the indispensable synergy of both modules in minimizing point reconstruction errors and generating high-fidelity continuous surfaces. Finally, real-world application on a crab claw point cloud demonstrates that the algorithm can effectively eliminate complex noise while maintaining feature integrity.
Keywords:
point cloud processing
improved statistical denoising algorithm
noise probability
KD-tree
mean shift
Journal
O
IF:
1.2
Papers:
178
Citations:
1.1W
Organization
Cited Papers