Extraction of Image Domain Primitives with a Network of Competitive/Cooperative Processes
The central goal of the first step in the visual processing of images is the detection and localization of image domain primitives. Such primitives along with their significant groupings should act as precursors to the recognition of the threedimensional structure of a viewed scene. Based on mathematical considerations as well as on results of studies in neurophysiological research, a framework for two-step filtering for feature detection has been developed. This allows for the detection of different kinds of linear features such as edge and line primitives (e.g. bars, slits). In order to reduce positional as well as directional uncertainty, further processing of the results from the initial filtering steps is necessary. With the aid of a competitive/cooperative processing architecture, which is an extension of the GROSSBERG/MINGOLLA scheme, gaps within the distribution of aligned operator responses can be closed and the detection of higher order primitives (such as junctions) is supported. Furthermore, as part of this architectural concept, a simple “winner-toke-all” inter-scale processing scheme has been realized. Simple grouping phenomena (based on spatial proximity of primitives) as well as vector fields (aligned with the structure of isophotes in homogeneous image regions) can be extracted.
Unable to display preview. Download preview PDF.
- /l/.D. Marr, E.C. Hildreth. 1980. Theory of Edge Detection. Proc. of the Royal Society of London, Vol.207, Series B, pp. 187–217Google Scholar
- /4/.V. Torre, T. Poggio. 1984. On Edge Detection. MIT, AI Laboratory: AI Memo 768Google Scholar
- /5/.A. Korn. 1988. Towards a Symbolic Representation of Intensity Changes in Images. Preprint, to be published in IEEE Transactions on Pattern Analysis and Machine IntelligenceGoogle Scholar
- /6/.W.E.L. Grimson. 1981. From Images to Surfaces: A Computational Study of the Human Visual System. Cambridge: The MIT PressGoogle Scholar
- /9/.H.-H. Nagel. 1985. Principles of (Low–Level) Computer Vision. Proc. Conf. on Fundamentals in Computer Understanding: Speech, Vision, Natural Language, Versailles, France, May 28–June 7 (preprint)Google Scholar
- /13/.H. Neumann. 1988. Theoretische Untersuchungen zur Extraktion monokularer Tiefenhinweise (Konturen und Schattierung) und ihre partielle methodische Evaluierung in einem rechnergestützten Perzeptionslabor. Universität Hamburg, Fachbereich Informatik: Dissertation (in print)Google Scholar
- /15/.A.P. Pentland. 1982. Local Analysis of the Image: Limitations and Uses of Shading. Proc. IEEE Workshop on Computer Vision: Representation and Control, Ringe, USA, Aug.23–25, pp.153–161Google Scholar