Images of bad quality require robust, global techniques for object recognition. The Hough technique for detection on curve-like image characteristics is well known. Parallel hardware, alternative data structures and projection techniques increase the efficiency of such methods, but the problem of interpreting the resulting accumulator remains. A backmapping technique is discussed which reduces the complexity of evidence information in accumulator space, and uniquely links specific points in image space to most-evident locations in accumulator space. The interpretation is substantially simplified and the accuracy of recognition increased. The global grouping of image points through common relations to evident accumulator cells offers the possibility of applying much more complex interpretation strategies. The advantage of the technique is illustrated by applying the method to three completely different recognition tasks in biomedical and medical imagery, where the link between points in image space and accumulated evidences is used in different ways.
|Original language||English (US)|
|Title of host publication||Unknown Host Publication Title|
|Number of pages||6|
|State||Published - 1987|
ASJC Scopus subject areas