ALGORITHMS FOR IMAGE REGISTRATION, MOVING OBJECT DETECTION AND IDENTIFICATION
Abstract
Image processing and analysis at video rates is a problem of wide interest. One way to perform these computations at a high data rate is to use a special parallel computer architecture. In this case, parallel algorithms must be developed which can be efficiently implemented on the parallel hardware. The research presented here is concerned with the development of fast, parallel algorithms for both the image processing and object classification phases of a real-time analysis task. Algorithms for processing include a parallel fast median filtering, pseudo median filtering, and a fast registration technique. The application of registration for temporal enhancement of the image and real-time tracking and detecting of moving objects are presented. The second phase of this research deals with shape analysis and object classification through the use of normalized moments. A new method of moment normalization has been developed to improve classification accuracy. These moments provided a set of meaningful global features which are independent of position, orientation, size and illumination of the object they represent.
Degree
Ph.D.
Subject Area
Electrical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.