Detection and tracking performance with compressed wide area motion imagery

Abstract

The current generation of digital sensor platforms used in remote sensing and surveillance provide a wealth of wide area motion imagery (WAMI). WAMI is employed in a variety of applications, including target detection and tracking. However, airborne platform limitations require that WAMI must be compressed to be effectively transmitted and stored. One standard image compression coding system commonly employed with WAMI data is JPEG2000. In order to meet platform and application requirements, JPEG2000 is often used in lossy mode and at high compression rates, leading to serious image degradation. This work studies the effects of image compression on the performance of the Air Force Research Laboratory's (AFRL) Government Algorithms for Tracking Exploitation Research (GATER) tracking algorithm at compression rates of 10:1 to 80:1. Additionally, several image enhancement algorithms widely available in literature, including Contrast-Limited Adaptive Histogram Equalization (CLAHE), High Frequency Boosting and Multiple Scale Retinex (MSR) and a novel approach titled Locally Tuned Sine Nonlinearity (LTSN) are applied to the imagery prior to compression in an effort to improve tracking performance. Results indicate the GATER tracking algorithm is measurably resistant to the effects of image compression up to 40:1 compression. At higher compression ratios, the algorithm is still able to detect the moving objects, however, false alarms become the most performance limiting factor. The enhancement algorithms performance is scene dependant; in some cases boosting overall performance and in other cases showing measurable performance degradation.