Assessing the performance effects of data compression for SAR imagery

Abstract

Although there are many ways to assess the impact of compression on image quality, in the final analysis one of the most important metrics is the impact of compression on end-users. This study evaluated compression effects on manual exploitation performed by military imagery analysts.. The original SAR imagery and compressed products were rated for radar NIIRS, image features, sensor artifacts, and target detection and recognition. JPEG, wavelet/trellis-coded quantization (W/TCQ), and intelligent bandwidth compression (IBC) were tested at 50 and 100-to-1 ratios. This study found that the utility of the compressed imagery's utility differed only slightly from the original, except for the JPEG products. Otherwise, both 50 and 100-to-1 compressed imagery appeared similar in quality. Radar NIIRS indicates that even 100-to-1 compression using IBC or W/TCQ has minimal impact on imagery intelligence value. A slight loss in performance occurred for vehicle counting and identification tasks. These findings suggest that both single scale IBC and wavelet/TCQ compression techniques have matured to a point that they could provide value to the tactical user. Additional assessments may verify the practical limits of compression for SAR data and address the transition to a field environment.