A Study on Seeded Region Based Improved Watershed Transformation for Brain Tumor segmentation Mahua Bhattacharya 1 , Arpita Das 2 1 Indian Institute of Information Technology & Management, Gwalior Morena Link Road, Gwalior-474010, India (Telephone: 91-0751-2449828, +919329550147, email:mb@iiitm.ac.in) 2 Institute of Radio Physics & Electronics, University of Calcutta 92 A.P.C. Road, Kolkata-700009, India (email: dasarpita_rpe@yahoo.co.in) Abstract The objective of segmentation is to partition an image into regions. In present work authors have presented an innovative image segmentation technique based on mathematical morphology using watershed transformation. Present approach is an intuitive method that produces a complete division of the image in separated regions avoiding the need for any kind of edge linking. We have further proposed seed-region based improved watersheds which removes drawbacks of conventional watersheds utilizing the prior knowledge of the test images. 1. Introduction Segmentation subdivides an image into its constituent regions or objects. The level to which the subdivision is carried depends on the problem being solved. The process of segmentation should stop when the objects of interest in an application have been isolated [1]. Medical image segmentation is one of the difficult tasks in image processing and accuracy of segmentation of region of interest ROI from an medical image determines the eventual success or failure of proper diagnosis. In this context of segmentation of medical images authors have developed a robust technique for segmentation using watershed transform using mathematical morphology [2]. We visualize an image with two spatial coordinates and gray levels and to understand the watershed transform let us imagine that there is a hole in each local minimum and that the topographic surface is immersed in water. Water starts filling all catchment basins at a uniform rate. If two catchment basins would merge as a result of further immersion, a dam is built all the way to the highest surface altitude and the dam represents the watershed lines. Final dam represents the continuous boundaries extracted by watershed segmentation algorithm [3]. It produces a complete contour of the images and avoids the need for any kind of contour joining. To avoid oversegmentation [4] some pre or post processing methods have been proposed in order to produce a more reasonable segmentation that reflects the layout of objects [5],[6-8],[9]. Local variations of the image can change the results immensely. Due to noise and other local irregularities of the gradient, some times watershed transformation produces fatal effect. This effect can be removed by blurring the images, which suppresses the noise and other local irregularity [10]. In proposed method, we have implemented marker-based improved watershed algorithm for segmentation of tumors appearing in human brain where seeded region-growing method has been used as the marker of the region of interests. In present paper, we have set the seed point on the basis of prior knowledge of the tumor regions with a consultation of the physician. 2. Algorithms of Improved Watershed Transform Our proposed method is based on improved watershed transform. In order to avoid oversegmentation images are being blurred. Then markers have been used to achieve meaningful segmentation. This marker is selected on the basis of seeded region growing method. Proposed watershed algorithm has been implemented on brain images having space occupying tumor lesions in order to segment the tumors from normal brain tissues. Each of the steps has been discussed in details.