图像中非凸区域的粗糙关系特征研究

Relation Character Research about Non-convex Regions in Image

通过研究图像中非凸区域的特性和A Cohn提出的"蛋-黄"模型,定义最大可能凸子集和最小可能凸超集等概念,给出了非凸区域的上、下近似概念,在此基础上提出了一种非凸区域转化为凸区域的粗糙近似算法,然后研究了非凸区域间、凸区域间的关系特征,给出了非凸区域间的粗糙关系与凸粗糙区域间的粗糙关系等价图,从而简化了非凸区域之间的关系。这对基于图像的空间定性推理(QSR)进行了推广,使其不但能够对含有凸区域的图像进行语义推理,而且能够对含有非凸区域的图像也能进行推理。

Via studying characters of non-convex region in image and ‘egg-yolk＇ model proposed by A Cohn, defining Max-Probility Convex subset, Min_Probility_Convex superset concepts, and giving high-approximation and low-approximation of, a rough approximation algorithm is presented in this paper that non-convex regions is converted to convex regions. Then through studying the relation properties among non-convex regions and convex regions, and giving a equivalence rough relation between non-Convex region and convex region. This work can generalize QSR based on image, and not only be suitable for semantic reason in image that contain convex regions, but also be suitable for semantic reason in images that contain non-convex regions.