Zriakhov M. Methods of preliminary processing and automatic adaptive compression of multichannel data in remote sensing systems

Research object – multichannel 1D signals of flying vehicle control systems, multichannel and one-channel images formed by remote sensing (RS) systems. Research goal – design of automatic compression methods that allow accounting statistical peculiarities of multichannel signals and images formed by RS systems. Research methods – processing of stochastic processes, spectral-correlation analysis, nonlinear filtering and robust estimation of noise type and level, methods of RS image classification, mathematical simulation, coding theory. Novelty – the method of multichannel signal compression is advanced, it is based on the use of 2D DCT and preliminary processing operations that allow increasing compression ratio due to increasing the similarity of input data; the method of selecting the parameter that controls compression is advanced, it provides user pre-set quality of decompressed data due to iterative search; in the first time the automatic method of compressing images corrupted by additive and signal-dependent noise in the neighborhood of optimal operation point is designed; in the first time the automatic method for compressing preliminarily filtered images is designed, it is based on estimation of residual noise variance after filtering; in the first time the method for automatic compressing of multichannel RS image (like the images of the systems AVIRIS, Landsat TM, etc) is designed, it is based on 3D DCT in groups formed adaptively with taking into account noise characteristics in channels. The proposed methods are realized as software and implemented in several enterprises. Application area – compression of multichannel RS signals and images.