Research On Ultraspectral Atmospheric Infrared Remote Sensing Images Compression Based On Key Information Preserving

The ultraspectral atmospheric infrared remote sensing images are radiance data obtained by observing the earth from an atmospheric vertical sounder mounted on a meteorological satellite.They generally have thousands of spectral channels in the infrared segment with extremely high spectral resolution and cover a wide range of space.The massive data generated daily have brought great challenges to transmission and storage.Therefore,it is very necessary to carry out the research on the compression of ultraspectral atmospheric infrared remote sensing images.In terms of application,the ultraspectral atmospheric infrared remote sensing images carry abundant information on the state of the earth’s atmosphere and are widely used in many fields,such as numerical weather prediction and climate research.It is neither necessary nor optimal to use all the channels in the retrieval process as the information content of these channels is highly redundant and the huge amount of data bring computational burden.Radiance data need to be thinned before entering the assimilation system to obtain the key subsets data,which should be protected.Therefore,An ultraspectral atmospheric infrared remote sensing images compression method based on key information preserving is proposed.Firstly,compared with the traditional hyperspectral images,the imaging characteristics,physical properties,data application processes and data redundancies of ultraspectral atmospheric infrared remote sensing images are analyzed.The data characteristics of AIRS ultraspectral atmospheric sounder are also introduced.Then,a key information extraction scheme for assimilation application and compression is proposed.The scheme includes a channel selection method based on information capacity and correlation constraint and a space sampling method based on information entropy,which is used for extracting the key data subsets data from two aspects of spectrum and space.A one-dimensional variational assimilation is performed and the calculated temperature profile obtains an effect close to the true sounding data.In addition,in the key information extraction process,the information capacity and the correlation constraint are taken into consideration,which facilitate further compression with taking into account the application requirements.Finally,a compression and decompression scheme based on key information preserving are designed and implemented.In the process of compression,first,a lossless compression is performed on the key subsets data using a compression method based on principal component analysis.Then,the key subsets data are used to predict the total data using the reconstruction prediction method for spatial and spectral dimension,and then the prediction residual data are further processed.Finally,the set partitioning in hierarchical trees coding and the range coding algorithm are respectively used to encode the prediction residual data in lossy and lossless to obtain the compressed bit streams.At the process of decompression,the total data can be restored using the opposite processes of compression.The standard ultraspectral atmospheric infrared data acquired by the AIRS ultraspectral atmospheric infrared sounder are used for compression.The experiment results show that,in lossy compression,the compression ratio decreases as the peak signal-to-noise ratio(PSNR)increases.With the PSNR greater than 40 d B,the highest compression ratio can reach 3.74.In lossless compression,the compression ratio can reach 2.81.The compressed bit streams focus on the protection of the key subsets data.The amount of data is reduced,and it is conducive to data processing and assimilation calculations.