Fuzzy graph modeling for text segmentation from land map images
Samit Biswas, Amit Kumar Das · 2012
Map image text segmentation has always been one of the difficult tasks because of its variety. The texts in a map may have the myriad background consists of various intensity values, different orientations, overlapping objects, intersected lines etc. Common problems for text extraction from map images are the lack of prior knowledge of text features such as color, font, size and orientation as well as the location of the probable text regions. Extracted texts can be used as an input to OCR for recognition. This paper presents an approach for text segmentation from map images using fuzzy graph analysis. Fuzzy graph is constructed from the map image. Fuzzy similarity value between two nodes within text region will be higher than other non-text regions. Seed points are selected through the fuzzy graph analysis. These seed points lie within texts in a map image. F* seed growing algorithm is used here for text localization.