If you were a soldier fighting in a dense jungle or pinned down by a desert sandstorm, you would no doubt like to receive real-time images of the battlefield, including pictures of tanks, trucks, and other targets.
But such images, say transmitted from a Global Hawk or other unmanned aerial vehicle (UAV), exceed the bandwidth of narrow communications channels that soldiers use.
Image compression algorithms, however, squeeze out unnecessary data in images, and might allow soldiers, including infantry and armour units, to receive battlefield information in real-time.
But would these compression algorithms leave images unclear even to highly trained analysts?
To find out if that was the case, Dr. John Irvine, from the US-based Science Applications International Corporation (SAIC) analysed the impact of several image compression techniques, including intelligent bandwidth compression (IBC), wavelet/trellis-coded quantisation (which SAIC and the University of Arizona developed), and the Joint Photographic Experts Group (JPEG) algorithm.
Dr. Irvine’s study focused on the ability of military imagery analysts to extract intelligence from synthetic aperture radar imagery compressed at ratios of 50-to-1 and 100-to-1 by these techniques.
The analysts viewed several renditions of each image in pseudorandom order – the original uncompressed image was always last – to try to identify background features (streams, vegetation, rocks, and roads, etc.) and military targets, such as tanks, armoured personnel carriers, and trucks.
Dr. Irvine found that tactical synthetic aperture radar imagery can be compressed significantly and still yield virtually all intelligence value. This suggests that both the intelligent bandwidth compression and wavelet/trellis-coded quantization techniques have matured to a point where they could provide value to the tactical user, such as a soldier on a battlefield.
(JPEG performed poorly at 50-to-1 ratio and was not used at 100-to-1 compression ratio.)
In addition, if the compression techniques were coupled with improvements in image registration (matching two images so that they correspond to the same physical region of the scene being imaged), battlefield commanders could use this information to make targeting decisions.
The article that Dr. Irvine co-authored with Richard Peters and Rhonda Ritzel, “Evaluation of the Tactical Utility of Compressed Imagery,” appeared in Optical Engineering.
The work was sponsored by the Defense Advanced Research Projects Agency.