A Denoising Method of Hyperspectral Remote Sensing Image Based on Wavelet Transform

Since different types of noise are inevitably introduced in the processes of image formation and transmission, image denoising is a necessary pre-processing process before various image applications. In this paper, a local adaptive wavelet denoising method based on elliptic direction window and edge detection is proposed. The method first performs wavelet decomposition for the image and performs edge detection on the wavelet coefficients. Then, the wavelet coefficients of the image are sampled by the elliptic directional window, and the local threshold of it is calculated. Next, the wavelet coefficients are quantized by the soft threshold function. Finally, the denoised image is obtained by inverse wavelet transformation. In addition, it is noted that weight less than 1 is multiplied to reduce the threshold amplitude as much as possible to preserve the edge features of the image. To validate the performance of the proposed denoising method, four standard gray-scale test images and hyperspectral remote sensing images are employed and the denoising results are compared with the Local Wiener Filtering with Directional Windows (LWFDW). The experimental results show that the method proposed in this paper performs better in the numerical indicators of classification of the hyperspectral image, and has fewer pseudo-Gibbs phenomena in visual than the LWFDW.