Acquisition And Reconstruction Algorithm Of Multispectral Images

The traditional color reproduction based on surface color of the tristimulus values, there ismetamerism and other issues, which can not get the consistent reproduction of the color ofobjects under various viewing conditions, it is difficult to truly achieve the goal of "WYSIWYG",so the spectrum-based color reproduction research came into being. How to accuratelyreconstruct the spectral reflectance from the known surface color of the digital value of thelighting conditions, It is an important direction for further study in the field of color images.To select a suitable spectral reconstruction experimental samples, the paper select ColorChecker24color standard color card as the research object. Thesis in experimental dataacquisition to build two sets of spectral data acquisition system, namely scientific grade CCDmonochrome multi-channel multi-spectral camera systems and Hyper Spectral Analyzer System.This paper introduces two common spectral reconstruction algorithm, the pseudo-inversemethod,and principal component analysis, with pointing out that the problems are based on theraw spectral data of the training sample, while ignoring the test image color characteristics. Thusthe introduction of two new spectral reconstruction algorithm, which based on the original imagecolor digital response value,namely R-matrix method and polynomial regression; they caneffectively reproduce the chrominance information of the original spectrum, but the reproductionaccuracy of the spectrum there inadequate. Therefore proposed a new and improved spectralreconstruction algorithm will be polynomial regression method with the R-matrix methodtogether to build a new method of spectral reflectance.The final text is the spectral recovery experiments for these five spectral reconstructionalgorithm, respectively, from the root mean square error, the color, the relative error and GFC, tocomparating analysis of the pros and cons of five reconstruction algorithms. Experimental resultsshow that a combination of improved R-matrix and polynomial regression algorithm is feasiblealgorithm for spectral reconstruction, and reconstruction algorithms are superior in coloraccuracy and spectral accuracy.This paper presents the research concept and technology path, given in-depth study of themulti-spectral image acquisition system with a new idea, hi-fi art, such as save, modify, restoreand replication is of great significance.