A novel thresholding method for automatically detecting stars in astronomical images

Tracking the position of stars or bright bodies in images from space represents a valuable source of information in different application domains. One of the simplest approaches used for this purpose in the literature is image thresholding, where all pixels above a certain intensity level are considered stars, and all other pixels are considered background. Two main problems have been identified in the literature for image thresholding-based star identification methods. Most notably, the intensity of the background is not always constant; i.e., a sloping background could give proper detection of stars in one part of the image, while in another part every pixel can have an intensity over the threshold value and will thus be detected as a star. Also, there is always some degree of noise present in astronomical images, and this noise can create spurious peaks in the intensity that can be detected as stars, even though they are not. In this work, we develop a novel image thresholding-based methodology which addresses the issues above. Specifically, the method proposed in this work relies on an enhanced histogram-based thresholding method complemented by a collection of auxiliary techniques aimed at searching inside diffuse objects such as galaxies, nebulas and comets, and thus enhance their detection by eliminating noise artifacts. Its black-box design and our experimental results indicate that this novel method offers potential for being included as a star identification module in already existent techniques and systems that require accurate tracking and recognition of stars in astronomical images.