A scheme for solving the object-background problem based on an object model of things to be recognized. The identification of boundaries in a scene being recognized is an important problem in image recognition. While solving this problem, a portion of the image containing only the object to be recognized must be distinguished. Then, only the features that are observed in this "object portion" of the image can be thoroughly described and analyzed. Complete analysis of the entire scene is costly in terms of time and computational resources, while the identification of an object using the features found only in the "object portion" of the image is expected to be easier because an analysis of such a portion does not have to take into account the background.
展开▼