In recent years, muti-spectral imaging technology has been widely used in some high-precision color reproduction area. Multi-spectral imaging technology effectively eliminated the phenomenon of metamerism and it can accurately capture and display color information. However, multi-spectral image data has numerous channels and large amount of data, which makes it difficult to store and transport them. How to effectively reduce the dimensions of multi-spectral images and the amount of data is a very important task.This paper made a detailed study of multi-spectral imaging and spectral reflectance estimation techniques, and proposed spectral reflectance estimation algorithm based on principal component analysis and independent component analysis. Then made a contrast of the result of spectral reconstruction from CIE1976 color, spectral reflectance Root-Mean-Square Error, RE and so on. Then it is experimentally proved that the ICA approach is better than PCA to reconstruct spectral reflectance. Base on the in-depth study of PCA-based dimensionality reduction of feature extraction, this paper proposed an ICA-based dimensionality reduction of feature extraction algorithms, and experimentally demonstrated. This algorithm can efficiently compress multi-spectral image data, and maintain the original spectral information of the image... |