Data Compression and Correlation Filtering: A Seamless Approach to Pattern Recognition

Mahalanobis, Abhijit, Lockheed Martin; Daniell, Cindy Evors, Hughes Research Labs.

doi:10.1117/3.PM91.ch6

Book: Smart Imaging Systems

Editor(s): Bahram Javidi

Author(s): Abhijit Mahalanobis, Cindy Daniell

Published: 2001

https://doi.org/10.1117/3.PM91.ch6

Author Affiliations +

Abstract

This chapter deals with correlation filtering techniques developed for finding patterns in wavelet compressed imagery. The work focuses on concepts in which image data is frequently stored and transmitted in compressed formats. It may be necessary to perform recognition functions at various points in a distributed image processing system that includes remote sensors, data links, and time-critical applications. Therefore, the problem of recognizing patterns in compressed images is an important one that has received some interest recently. The simplest strategy perhaps is the one shown in Fig. 6.1, where the image is first uncompressed and then the recognition algorithm is applied. It is, however, desirable to integrate the recognition and compression algorithm so that patterns can be found directly in the compressed data without the need for explicit image reconstruction. Nevertheless, it is useful to study the system in Fig. 6.1 to treat its performance and throughput requirements as a benchmark against which improvements can be assessed.