Year: 1993
Type: report
Abstract: In this paper, I review and discuss recent advances in the development of algorithms for solving the kinds of geometric matching problems that occur in computer vision, in particular under bounded error models. I examine both the one-model case (recognition) and the many-model case (indexing), and I discuss some open problems and directions for research. 1 Introduction Visual object recognition is concerned with finding instances of objects in images. There are many aspects to this task, but one of the most important aspects is geometry. That is, given a geometric model of the shape of an object, a recognition system is to determine whether some projection of that shape is (partially) present in an image. Visual object recognition has important applications in robotics, communications, and information retrieval. Because such applications usually have to cope with large amounts of image data, and since many of the problems in visual object recognition have a "combinatorial flavor", d...
Author: Thomas M. Breuel
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W560938031
