Author: A. A. Sirota, A.I. Solomatin, E.V. Voronova (Voronezh State University, Voronezh, Russia)
Computer Optics, Volume 34, Issue 1, Pages 109-117 ( January 2010)
Abstract
In this paper there is considered a two-staged object detection algorithm on images with random cutout shape and in case of addictive noise presence. On the first stage the local image parts are analyzed using statistically optimal or neural algorithms to detect and estimate the brightness jump parameters. On the second stage the final decision are made about object presence on the image and about its cutout integrity by analyzing the local parts initial processing results using maximum likelihood algorithm There is suggested an algorithm to increase object detection process. This algorithm finds the maximum of likelihood functional by searching a minimal path on the graph by dynamic programming.
Keywords
Brightness jump, image processing, neural networks, object cutouts selection, object recognition
Full text
EID: 2-s2.0-84938071665