2015 International Symposium of Computer Application and Information Technology (IsCait 2015)

290 / 1971-01-01 00:00:00

Multi-spectral Color Signal Classification By Means Of Modified Spectral Similarity Metrics

color signal,similarity metrics,spectrum angle,Fre ́chet distance

Abstract—Multi-spectral image has rich spatial and spectral information, in which color signal or each pixel treated as the discrete spectral curve. In order to achieve the multi-spectral image compression during transmission, color signals must be classified by the method of spectral similarity comparative. Defining the spectral distance and concerning the shapes of curves distribution, a new method for spectral similarity metrics was proposed using the spectral angle SAM and discrete Fréchet distance. Experimental results showed that the novel method can classify the multispectral image color signals effectively, and adapt to image noise well.