|M.Sc Student||Kligler Netanel|
|Subject||On Visibility and Image Processing|
|Department||Department of Computer Science||Supervisor||Professor Ayellet Tal|
In this work we give visibility detection within a point set (known also as a "point cloud") a new interpretation. What does it mean that a point was detected visible? Although this question has been widely studied within computer graphics, it has never been regarded when the domain is abstract and the point set consists of feature vectors (rather than a real scene). We show that this question is indeed relevant, even for the abstract case. Although this question may be relevant in many abstract domains, we focused our research to that of image processing. Specifically, we show that a simple representation of an image as a 3D point set lets us use visibility detection in classical image processing tasks. Given an image, each pixel is represented as a feature point. A viewpoint is set, and points that are visible to the viewpoint are detected. Utilizing the visibility information is later combined into existing algorithms, having the potential to improve state of the art results in a variety of applications in the field. To demonstrate this concept, we do not create new visibility-based algorithms. Rather, we alter existing ones. The existing algorithms, combined with the unique visibility information that we supply, have the strength to present better results, outperforming current state-of-the-art algorithms.
As proof of concept, we demonstrate this idea within three specific applications in image processing: text image binarization, document unshadowing and stippling-style illustration.