Indus Image Segmentation Using Watershed and Histogram Projections
Anupama Namburu · International Robotics & Automation Journal · 2017
Character segmentation is the major step of document image analysis and optical character recognition (OCR). The character segmentation is necessary to detect all the character regions in the image document. The proposed method preprocesses the image document with edge detection techniques to enhance the character edges. Further, the watershed algorithm is implemented to identify the regions of the character. Also, the multiple histogram projections are used to identify the characters. The watershed regions and the multiple histogram projections are compared to analyse the actual character regions improving the accuracy of character recognition. The Proposed method is evaluated on Telugu and Indus images and has extracted the characters accurately.