Mining textural association rules in RS image
Zuocheng Wang, Lixia Xue, Dejun Feng · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2007
Based on gray and texture features of remote sensing (RS) image, a new method of textural combined association rules mining is proposed in this paper. According to the spectrum features of pixels of image, all the pixels constructing the textural RS image and all the texture cells have relationships between each other. This is premise of mining association rules in image. In order to mine the textural association rules in RS image, each image can be seen one transaction, and frequent patterns can be mined. If image data mining drills down to pixel level, each pixel or its neighborhood can be seen one transaction too, and data mining was processed in all the transactions. In textural image, the frequent patterns are texture cells in fact. Because of different size of texture cells, multi-levels and multi-masks data mining was studied. Based on definition of image association rules, one association rule represents the local structure of RS image, and the support s% and confidence c% denote the possibility of the pattern. The experimental results validate that the combined association rules can represent the regular texture, and can represent the irregular texture perfectly too. By the combined association rules we can accomplish image segmentation.