COMPARATIVE ANALYSIS OF EDGE BASED AND REGION BASED ACTIVE CONTOUR USING LEVEL SETS AND ITS APPLICATION ON CT IMAGES

B.S. Saini, G.Sethi

Abstract: Detection of Liver cancer from CT images is an exigent task due to the reason that cancer impression in CT images are of very low in contrast and have indistinguishable edges. Indistinguishable edges are edges in which foreground and background are almost same. Image segmentation is used to extract desired anatomical structure from image. Image segmentation is the process of dividing the image into multiple regions. These regions are sometimes called region of interest (ROI). These ROI’s are used as informative inputs to further image processing e.g. feature extraction, selection and ultimately the classification of a disease. Thus an effective image segmentation is utmost important in medical images. In case of indistinguishable edges most segmentation techniques fails to detect edges. Edge based and region based active contour methods are most prevalent for image segmentation. Edge based technique works using gradient function and ultimately stopping function to detect edges , while region based uses average information inside and outside regions to control evolution and termination at edges. In this paper comparative analysis of edge based and region based active contour using level sets is done. When applying these methods on the CT images with an impression of liver cancer, it has been found that the edge based contour able to locate the desired edges more accurately. The quantitative and qualitative results comparison between two techniques has also been done. Results shows that edge based methods performs comparatively better than region based active contour using level set if number of iterations are controlled properly

Keywords: Level Sets, Gradient, Stopping function, Active contours and Liver cancer

DOI: https://doi.org/10.15623/ijret.2013.0204028