Segmentation of Objects in Digital Images Based on the Modified C-V Model

In this paper we introduce a scheme for segmentation of objects in digital images. The proposed method is based on the modification of the Chan-Vese (C-V) model that optimizing an energy function based on the homogeneities inside and outside of the evolving contour. A new regularizing term is added to the C-V model to increase the segmentation accuracy. The proposed method is tested on medical and non medical images for efficiency assessment. For medical images, the segmentation efficiency is analyzed quantitatively via area overlap ratio (AOR) between computer segmentation and manual segmentation by expert radiologist of mammographic masses selected from Digital Database for Screening Mammography (DDSM) database. At an overlap threshold of 0.5, the proposed method correctly segments 91% of the masses, while CV method delineates only 77% of the masses. For non medical images, the performance is assessed qualitatively via visual inspection of various images selected from Berkeley segmentation dataset. Hence, the technique presented here has shown a very encouraging level of performance for the problem of object segmentation in digital images.