Efficient encoding and decoding algorithms for distributed resampling

Abstract

This paper discusses algorithms for compressing floating point data which is encountered while performing resampling algorithm for rectifying geometric distortions on the images transmitted by the remote sensing satellite on a distributed computing environment. Two efficient algorithms for encoding and decoding floating point data compression, (i) 3-byte packing and (ii) 3-byte + 2-bit packing have been proposed. Both algorithms are tested to execute resampling algorithm on radiometrically corrected for IRS- LISS-III 4 bands data on a distributed system. First algorithm compresses 4-byte floating point data to 3-byte obtaining 25% compression while later one compresses 4-byte floating point data to 3-byte + 2-bit achieving about 18.75% compression. The computational time is reduced by 22% as compared to the distributed resampling algorithms without compression. Further it is found that in lossy compression algorithm only 220 pixels out of 37.4 MB pixels have utmost one Gray count difference, which will not pose any issues for digital classification or any other methods that will be employed in the corrected image by Image Analysts.