摘要
针对视觉任务中普遍存在的自遮挡现象和目前视觉系统尚无明确的自遮挡检测算法的现状,提出了一种结合视觉目标深度图像和最佳分割阈值迭代的自遮挡检测算法。在分析图像阈值分割技术的基础上,将分割阈值迭代法的思想引入深度图像领域,并结合使用视觉目标对应的深度差值图像信息,通过求取合适的阈值实现了对自遮挡现象的检测。实验结果表明,该方法能够有效地检测出视觉目标中存在的自遮挡现象并定位自遮挡边界,弥补了目前自遮挡检测领域研究的不足。
Aiming at the familiar serf-occlusion phenomenon in vision tasks and considering the actuality that there are no specific self-occlusion detecting algorithms, for vision systems, the paper proposes a self-occlusion detection algorithm combing depth image and optimal segmentation threshold iteration. At first, the segmentation threshold iteration method is applied to depth image by analyzing the threshold segmentation technology. Then, the self-occlusion detection is realized by obtaining a proper threshold and using the depth difference image information of the vision object. The experimental results show that the proposed algorithm can detect the serf-occlusion phenomenon and obtain the occlusion boundary effectively, so the deficiency in self-occlusion detection domain is compensated.
出处
《高技术通讯》
EI
CAS
CSCD
北大核心
2010年第7期754-757,共4页
Chinese High Technology Letters
基金
863计划(2006AA04Z212)
河北省自然科学基金(F2007000423
F2010001276)资助项目
关键词
深度图像
深度差值图像
阈值选择
迭代
自遮挡检测
depth image, depth difference image, threshold selection, iteration, self-occlusion detection
