Removing the stripe noise in HJ-1A HSI images

Abstract

The first 20 channel images of Hyper Spectral Image (HSI) have obvious stripe noise, which seriously affects the application of the HSI data. This paper summarized and analyzed the existed destriping methods. Based on the image characteristics of HSI, a new stripe noise removal method was proposed by using spline interpolation function to resample. Taking the HSI image of Guangzhou area as an example, the Fourier transform method, moment matching method and the new method proposed in this paper were used to remove the stripe noise. The results showed that the Fourier transform method couldn't remove the image stripe noise effectively; the moment matching method could remove the image stripe noise, but it also forced the image mean column value to 1, which would destroy the differences of image features information; The new method proposed in this paper was able to retain the differences of image surface features while removing the image stripe noise. The destriping efficiency of the new method was better than the other methods.