论文部分内容阅读
提出了一种能够良好地保持高光谱遥感图像细节特征的噪声去除方法。该方法首先利用噪声调整的主成分分析(NAPCA)进行特征提取,再利用复小波变换(CWT)对NAPCA变换后的低能量成分进行去噪处理。对此低能量成分的每个波段利用二维复小波去噪,此时复小波系数采用Biva Shrink函数进行收缩。然后对低能量成分的每条光谱进行一维复小波变换,利用邻域阈值函数进行小波系数的收缩。对AVIRIS图像贾斯珀桥、月亮湖和盆地进行的仿真实验表明:该方法去噪后的信噪比与HSSNR相比提高了4.3~7.8 d B,与PCABS相比提高了0.8~0.9 d B,验证了该算法的可行性。真实数据OMIS图像的实验结果验证了该方法的有效性和适用性。
A noise removal method that can maintain the detail characteristics of hyperspectral remote sensing images is proposed. Firstly, feature extraction is performed by principal component analysis (NAPCA) of noise adjustment, and then the complex wavelet transform (CWT) is used to denoise the NAPCA transformed low energy components. For each band of low energy components, two-dimensional complex wavelet denoising is used. In this case, the complex wavelet coefficients are contracted by the Biva Shrink function. Then one-dimensional complex wavelet transform is performed on each spectrum of the low-energy components, and the wavelet coefficients are contracted by using the neighborhood threshold function. Simulation results on the AVIRIS image of Jasper Bridge, Moon Lake and Basin show that the signal-to-noise ratio improved by 4.3-7.8 d B compared with HSSNR and 0.8-0.9 d B compared with PCABS , Verify the feasibility of the algorithm. The experimental results of real data OMIS image verify the validity and applicability of this method.