Mining frequent spatial patterns in image databases

Mining useful patterns in image databases can not only reveal useful information to users but also help the task of data management. In this paper, we propose an image mining framework, Frequent Spatial Pattern mining in images (FSP), to mine frequent patterns located in a pair of spatial locations of images. A pattern in the FSP is associated with a pair of spatial locations and refers to the occurrence of the same image content in a set of images. This framework is designed to be general so as to accept different levels of representations of image content and different layout forms of spatial representations.