Skip to Main Content
Region selection problems from raster images in Geographic Information Systems (GIS) involve lots of computational complexity for which designing of the optimized algorithms is required. A region with maximum sum of an attribute value is useful in different decision making problems, for which no efficient solution has yet been found. For selecting a region of such interest, there are lots of unique regions in both 4 and 8-Adjacency region rules. In case of 4-Adjacency rule, for selecting a 10-cell region, more than 288 thousand unique regions can be found among one will be a region with maximum sum value. The scenario suggests that the problem falls in the class of NP-Hard decision problems, so the problem is complex in nature as we deal with the computation terms of the algorithm to be designed in polynomial time. This research contains a new dynamic programming approach for the problem, whose focus is on the efficiency for finding approximate solutions. Experiments have suggested that with the data input of images having higher degree of spatial auto correlation, the algorithm is quite useful and practical in nature, as it was tested with real life data and has extracted maximum value regions with thousands of cells in a sufficient time.