A convex formulation for color image segmentation in the context of passive emitter localization

In many tasks in information fusion objects of interest need to be extracted from color images. Often the only available information is the color of a specific object. In this paper we present a novel method for segmenting images into two regions, foreground (e.g. object) and background. We introduce a convex energy functional based on total variation that is subsequently solved using the Euler-Lagrange theorem and a parallel implementation of successive over-relaxation. The main achievement of this formulation is that, due to the convex formulation, the algorithm is guaranteed to find the global optimum from all possible solutions. Furthermore, this algorithm can be heavily parallized using the graphics processing unit (GPU). In the following we will show how to use the thus obtained image information in the context of passive emitter localization from aerial platforms. It will be shown that the fusion of image-and bearing-based localization results can strongly improve the bearings-only results.