From Primal Sketch to 2.1D Sketch

One of the principal goals of low-level vision is to segment an image. In a scene, maybe there are some objects and overlapping between different objects. As a result, it is impossible to segment integrated objects because of the unseen parts. So it is unacceptable. Therefore, we should use some information, such as primal sketch, color comparability, orientation comparability etc. to research how to layer different objects and connect the different parts of the same object. This is the main task of the 2.1D sketch research.

This research mainly care surface objects and their partial order and without regard to their deep information.

This research can be used to image segmentation, photos edition and motion etc.