Data compression techniques for underwater imagery

The onboard data storage burden associated with underwater (UW) imagery tends to be significant for applications such as image-based navigation via an autonomous UW vehicle (AUV). Due to mission, space, and power requirements the use and storage of multispectral imagery in compressed form is preferred. In this paper, the authors discuss the spatial and statistical characteristics of UW imagery that facilitate compression by well-known algorithms such as JPEG, vector quantization (VQ), and visual pattern image coding (VPIC). For example, they consider statistical distributions of target and background greylevels obtained from truthed imagery, as well as power spectral analysis of target-background differences. The former measures facilitate parameter selection in VQ and and VPIC, while the latter are important in JPEG. Preliminary results are given for recently-developed algorithms that yield compression ratios ranging from 5,500:1 to 16,500:1 based on prefiltered six-band multispectral imagery of resolution 720/spl times/480 pixels. The prefiltering step, which removes unwanted background objects, is key to achieving high compression.