Scalable compression of multibeam echo sounder data

We propose a scalable lossy-compression method to enable acoustic data transmission through a narrow band channel (e.g. from an AUV), while still allowing for sufficiently good reconstruction on the receive station, which is adequate for other processing applications. Scalable coders are well-known in regular image compression, but they have not been applied yet to sonar backscatter data from multi-beam echo sounders (MBES). We achieve so by converting the backscatter samples acquired during a survey into one waterfall-image, which is an optimized representation of the MBES backscatter data. Each line of the image is assembled based on an auxiliary ping time-series obtained by consolidating the individual beam time-series into a single signal, spanning the space of all the responses of the multiple overlapping beams. Then, the ping time series are stacked, forming a monochrome image aligned by the center of the navigation, without any geodesic referencing. The assembled waterfall-image is then compressed using standard image compression algorithms. The distortion metric we use to evaluate quality is the mean absolute error (MAE) of samples among original and reconstructed ping time series as well as the MAE among the original multi-beam time series. However, the most important distortion metric in our opinion is the comparison of compositions of a mosaic image of the sea bottom surface among compressed and uncompressed data. Results indicate compression ratios of up to 200:1 at a lower but useful quality, while smaller ratios may yield mosaic images from compressed data that are virtually indistinguishable from the original. Thus, we have a scalable compression method that allows for a large range of compression ratios. From the highly compact preview version that an AUV could send to the surface to the accurate data that can be near-losslessly stored for future studies. The proposed algorithm is efficient and we are unaware of any similar work in the literature.