Gradient based polyhedral segmentation for 3-D range image

Segmentation is a key step in vision applications. In this study a novel method is proposed to achieve robust and low-cost polyhedral segmentation for 3-D range image. A two-dimensional gradient histogram space is generated based on gradients along the directions of the x and y coordinates. The problem of surface segmentation becomes a problem of points clustering in the feature space. A four-neighborhood iterative expanding algorithm is proposed for region grouping according to the gradient feature space. For noise and geometrical distortion regions, a merge process is applied to the initial segmentation results. The selection of parameters used in this segmentation approach is also discussed. The experimental results show that the proposed algorithm provides a new approach of range image segmentation with the characteristics of low computational complexity and less sensitivity to noise.