Research On Spectral Reflectance Compression And Reconstruction

Spectral images have a wide application prospect in the field of high-fidelity color representation and reproducibility.Spectral reflectance is a unique identifier to characterize the color of an object,and its spectral image can accurately store the basic information about the color of an object,and spectral reproduction can achieve accurate color matching under any light source and standard observer.However,spectral data information is composed of many numerical values,which occupies a large amount of storage space and affects the data transmission speed.Therefore,how to compress and reproduce the spectrum has become a research hotspot.In addition,the original spectral space is not suitable for spectral image processing,color gamut mapping and color gamut boundary description.Therefore,an interim connection space(ICS)needs to be defined for the compression and reproduction of spectral images.At present,the interim connection space includes Lab PQR,XYZLMS,Lab RGB,etc.With the changing needs of the times,it is required to reconstruct the spectral reflectance with high spectral accuracy and chroma accuracy in the practical applications of cultural relic storage and remote sensing,etc.How to design the interim connection space that can improve the reconstruction accuracy is the main research objective of this thesis.In this thesis,three new interim connection spaces are proposed.Firstly,inspired by Zhang Xiandou et al.,Tsutsumi et al.,an interim connection space based on nonlinear programming is proposed.The compression matrix is modeled as a variable,and the relevant reconstruction matrix is selected as the "Weiner estimation matrix".The compression matrix is determined by solving the nonlinear minimization problem based on the training data set.Later,inspired by the Lab PQR interim connection space proposed by Derhak et al.,the mapping matrix in the spectral decomposition formula was selected as a "Weiner estimation matrix" which was not only affected by the lighting environment,but also depended on the training reflectivity data set to further improve the reconstruction accuracy.Finally,the interim connection space and spectral reconstruction method based on chroma value are proposed.The interim connection space consists of two groups of three stimulus values under two real light sources.Based on this new interim connection space,the mapping matrix in the spectral decomposition formula is selected as "Weiner estimation matrix",and then the spectral reconstruction method is improved,m important basis vectors for the metameric black space based on the new spectral decomposition,and a mapping matrix via a polynomial model of order k,were trained so that the reconstructive accuracy can be further enhanced.In this thesis,Munsell data set was used as the training sample and NCS data set as the test sample.The spectral accuracy was evaluated by calculating the goodness of fit coefficient and root mean square error of the reconstructed spectral reflectance and the original reflectance.The chroma accuracy was evaluated by CIEDE2000 chromatic difference formula and compared with other studies.In order to evaluate the practical application of the proposed interim connection space,spectral image data sets from several common color laboratories were used for testing.Experimental results show that the three new interim connection spaces proposed in this thesis have improved spectral accuracy and chrominance accuracy to varying degrees,and can play an important role in the application of spectral image compression and reproduction.