Study On Sparse Characteristic Of Color Image And Its Applications

As the sparse representation of the signal can effectively capture the most essential features of the signal, it plays a crucial role in modern signal processing, such as signal acquisition, feature extraction, signal processing, compression, encoding, transmission, storage, and so on. In order to minimize the pressure on image capture, transmission and storage of color images(even spectral images) with huge amount of data, we need analysis and obtain sparse, which is very important and also attract focus at home and abroad.In this thesis, a study of sparsity analysis on color images and spectral images has been expanded which is based on the background above. We firstly combine local characteristics and global characteristics of the image from the perspective of spectral dimension and spatial dimension. And then, using the basic imaging principle and local adaptive sparse PCA method we analyzed the sparse feature of color images comprehensively.On CS theory, we applied sparse in demosaicking and designed a hexagonal grid of color filter array(CFA) for the fuzzy edges and zipper effect problem demosaicking methods currently prevailing, and design demosaicking model based on CS. Finally, tracking method base on orthogonal matching was used for full-color image reconstruction. Experimental result shows that this method can overcome the blurred edges and zipper phenomenon and effectively improve the reconstructed image quality better than other methods.In addition, based on a new compression spectral imaging technology, this thesis also applied sparse in NEO imaging and established a space sparse reconstruction optimization model to solve the problems of huge amount of observation data, difficulty of storing and transport and so on for existing NEO spectroscopic imaging method, with dual-channel spectral imaging frame coding template-based which fully used the space sparse feature of the celestial spectra images. In this model, we used two-step iterative shrinkage algorithm to reconstruct a high quality spectral images, and effectively reduces the amount of data observed during imaging.In summary, this thesis carried out the sparse color image characteristic analysis to establish the sparse model, and carried out useful research for typical applications. Experimental result shows that the proposed method for color(spectrum) image processing has some theoretical significance and application value.