Stripe Noise Removal from Remote Sensing Images Based on Stationary Wavelet Transform

Abstract

Fourier transform is applied to detect the direction of stripe noise before de- noising, which is advantageous for selecting the corresponding detail coefficients for threshold quantization after stationary wavelet transform. Depending on the direction of stripe noise, the corresponding detail coefficients contain stripe noise need to be removed, while retaining the approximate coefficients and other detail coefficients. The algorithm can remove stripe noise effectively and preserve original image details as much as possible. In addition, stationary wavelet transform is used for the first time in stripe noise removal from remote sensing images. Compared with the traditional de-noising algorithms based on discrete wavelet transform, stationary wavelet transform can better maintain the details of original image because it is redundant and shift-invariant. As shown in the experiments, the new algorithm reported herein performs better than discrete wavelet transform, mean filter and gauss filter, leading to good visual effects.

Keywords

• Fourier transform
• Stationary wavelet transform
• Discrete wavelet transform
• Stripe noise
• Image de-noising.