利用多种光谱线性变换及其融合的方法,减少波段间的信息冗余,提高光谱特征的可分性,并用实验证明了算法的有效性。
Multiple linear speitral components extracted by PCA, Fast-ICA and IM-ICA are selected and fused, in order to reduce the information redundancy and improve the spectral feature space. The proposed algorithm is implemented on the IKONOS images and the experimental results show obvious accuracy improvements.