A New Spaceborne Compression Approach for Remote Sensing Imagery

摘要

After analyzing advantages and disadvantages of these typical encoding methods: SPIHT and VQ, a "DWT+MRLE" approach for spaceborne data compression was proposed in this paper. This approach includes two steps: Discrete Wavelet Transform (DWT) and Modified Run Length Encoding (MRLE). The former used CDF9/7 biorthogonal wavelet filters to powerfully get rid of correlation between pixels in imagery. The later put enlightening information into the lowest bit of some key-position transform coefficients. Consequently, CDF9/7 and MRLE together make hardware platform remain high real-time capability, and help reconstructed images keep good fidelity with PSNR being about 40dB, compared with the original ones. Comparison between experimentations on SPOT4's low-spatial-resolution (10m) imagery and Ikonos2's high-spatial-resolution (1m) imagery, shows this "DWT+MRLE" method having better performance for remote-sensed imagery, especially those of higher resolution. Although inferior to 8:1, Compression Ration (CR) here near 5:1 is greater than France SPOT5's 3:1 and American Ikonos2's 11:2.6 on-board data compression. More important, this method having less computational amount is good for spaceborne capability of real time. The consumed time of different image size is also presented in this paper, based on TI TMSC6416 DSP chip with 600MHz CPU cycle clock.