Spectral Estimation Method Based On Ica Study

The color reproduction technique, based on the multi-spectral imaging and spectrum estimation, has become the front of the color science & engineering field since the nineties of last century. Because the spectrum estimation arithmetic needs to capture the spectrum reflectance of each pixel, the multi-spectral imaging system needs to capture and store more data than the three-channel RGB system. It is very necessary to extract spectrum feature and compress spectral data, that is, a large spectrum dataset can be represented by the linear combination of some eigenvectors. The principal component analysis is a kind of effective data compression method. However, the weights of the result of the PCA can only be irrelevance mutually, but not independently one another, except the original signals are Gaussian data.The independent component analysis is a kind of signal processing method, which has been developed since 90's, 20DC, and been widely used into the fileds of the Signal Feature Extraction, the Blind Original Signal Separation, the Physiology Data Analysis, the Finance Data Analysis, the Speech Signal Processing, the Image Processing and the Human Face Recognition etc. ICA doesn't been required the original signal is Gaussian. In recent years, the ICA is used to extract the feature of the spectrum data in the field of the color science.In this paper, the spectrum reflectance of 150 cases Munsell cards and 150 cases birch leaves are estimated, using the ICA method and the PCA method to extract the independent components and the principal components, separately, and to compress the spectral data in the spectrum estimation arithmetic. The estimation result of the ICA and the PCA is compared in thispaper.