基于法向分量边缘融合的深度图像分割  

Range Image Segmentation Based on Edge Fusion of Normal Component

作　　者：范剑英[1] 于舒春[1] 王洋[1] 于贵江[1] 于晓洋[1] 

机构地区：[1]哈尔滨理工大学测控技术与通信工程学院,哈尔滨150040

出　　处：《计算机工程》2010年第17期221-222,225,共3页Computer Engineering

基　　金：国家自然科学基金资助项目(60572030);黑龙江省教育厅科学技术研究基金资助项目(11541054;11541046)

摘　　要：针对深度图像边缘较难分割的问题,提出一种基于法向矢量分量边缘信息融合的深度图像分割方法。计算深度图像中每点的法向矢量,提取法向矢量的x、y分量并构建灰度图,融合Sobel算子的边缘检测结果得到深度数据的初始分割,通过细化处理得到最终分割结果。实验结果表明,该方法得到的分割区域边界闭合完整,分割质量较高。Aimimg at the problem that range image is difficult to segment, this paper presents a range image segmentation method based on edge information fusion of normal components. Normal vector of every point in range image is computed, these normal vectors are decomposed into x component and y component which are used in building gray images. It uses Sobel operator in detecting edges of every range image and then these edges information are fused into an initial segmentation, and gets the final result by thin processing. Experimental results show that segmentation region edge closes completely and segmentation quality is above normal.

关 键 词：深度图像分割 法向分量 边缘检测 信息融合 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]