Catalogue of Artificial Intelligence Techniques
Aliases: Resolution Cones
Keywords: array, processing cone
Categories: Pattern Recognition and Image Processing
Author(s): Bob Beattie
Strictly, a pyramid/resolution cone is a set of arrays used to store a digitised image at different resolutions, the highest resolution forming the base of the pyramid. Normally, there is a uniform relationship between adjacent levels of the pyramid, for example, a pixel intensity at a particular resolution is often the average value of the pixel intensities in the corresponding 2 x 2 pixel block at the next highest resolution. More generally, the concept of a regular hierarchical data structure has been extended to include processing elements within the structure, then known as a processing cone. Originally, pyramids were used to explicate the relationship between the different image resolutions in a computer vision system using planning. For instance, the boundaries found at a low resolution might be used to refine the Edge Detection/ Boundary Detection process at higher resolutions. This can significantly improve the efficiency of the overall edge detection process. More recently, it has been realised that image boundaries occur over a wide range of scales from sharp step-like edges to fuzzy, blurred edges. Some results have been stated on using the relationships between corresponding edges occurring at different resolutions to make assertions about the scene boundary producing them. Since these regular hierarchical structures are inappropriate for non-spatial symbolic computation their use is normally restricted to the early processing stage of a vision system.
- Tanimoto, S.L., Regular hierarchical image and processing structures in machine vision
(Hanson, A.H. and Riseman, E.M.
, eds.), Academic Press, New York, 1978.