An Improved Convexity Based Segmentation Algorithm for Heavily Camouflaged Images

Abstract

The paper proposes an advanced convexity based segmentation algorithm for heavily camouflaged images. The convexity of the intensity function is used to detect camouflaged objects from complex environments. We take advantage of arg D operator for the detection of 3D concave or convex graylevels to exhibit the effectiveness of camouflage breaking based on convexity. The biological motivation behind arg D operator and its high robustness make it suitable for camouflage breaking. The traditional convexity based algorithm identifies the desired targets but in addition also identifies sub-targets due to their three dimensional behavior. The problem is overcome by combining the conventional algorithm with thresholding. The proposed method is able to eliminate the sub-targets leaving behind only the target of interest in the input image. The proposed method is compared with the conventional arg D operator. It is also compared with some conventional edge based operator for performance evaluation.