This paper presents a novel method for image segmenting in thermal images. A thermal imaging sensor is a device that forms an image by detecting the infrared radiation from objects in a scene and comparing their temperature differences. In this study, we characterized an object's heat signature (its mean temperature and variance) by segmenting its contours in thermal images and analyzing the information from each foreground segment. However, because of the limitations of thermal sensors (e.g., a high signal-to-noise ratio and low contrast), manually contouring thermal images is difficult and time consuming. We propose a semiautomatic segmentation method that takes advantage of information from another optical camera. Such a bi-sensor system improves the precision of contour sketching for objects in thermal images. The experimental results demonstrated the robustness of the proposed method.