ارزيابي اثر نويز بر استحكام اندازه گيري شباهت طيفي

Assessment of Noise Effect on Robustness of Spectral Similarity Measure

Spectral similarity measure plays a key role in hyperspectral images processing through the spectral signature analysis. Since the conditions of providing reference spectra are not always the same and maybe affected by noise, therefore a method with the least sensitivity to noise should be applied for the spectral similarity measure. In this paper we discuss robustness of different spectral similarity measures with respect to noise which is affected on reference spectra. To measure the noise effect on robustness of spectral similarity measures, changes in value of spectral discrimination indices, including relative spectral discriminatory probability (RSDPB) and relative spectral discriminatory power (RSDPW) are investigated. Although all implemented methods have had a good performance in identifying the target with maximum discrimination power, after noise implementation; SAM method has shown the least sensitivity to noise within these two indices.