Shock-Based Reaction-Diffusion Bubbles for Image Segmentation

Abstract

Figure-Ground segmentation is a fundamental problem in computer vision. Active contours in the form of snakes, balloons, and level-set modeling techniques have been proposed that satisfactorily address this question for certain applications, but require manual initialization, do not always perform well near sharp protrusions or indent ations, or often cross gaps. We propose an approach inspired by these methods and a shock-based representation of shape. Since initially it is not clear where the objects or their parts are, they are hypothesized in the form of fourth order shocks randomly initialized in homogeneous areas of images which then form evolving contours, or bubbles, which grow, shrink, merge, split and disappear to capture the objects in the image. In the homogeneous areas of the image bubbles deform by a reaction-diffusion process. In the inhomogeneous areas, indicated by differential properties computed from low-level processes such as edge-detection, texture, optical-flow and stereo, etc., bubbles do not deform. As such, the randomly initialized bubbles integrate low-level information, and in the process segment figures from ground. The bubble technique does not require manual initialization, integrates a variety of visual information, and deals with gaps of information to capture objects in an image, as illustrated on several MRI and ultrasound images in 2D and 3D.¹.

The authors gratefully acknowledge the support of NSF under grant IRI-9305630.