A Striping Removal Method Based on Spectral Correlation in MODIS Data

Modis imagery suffers from nonlinear striping in many bands. The nonlinear stripes make great difficulties for conventional linear statistical destriping methods. This paper presents a two-step method for nonlinear stripe removing. Both steps depend on the correlation information between defective band and its reference normal band. First stripes in defective band are divided into different portions with the aid of its high correlation spectrum by GA algorithm automatically and moment matching destriping is performed in each portion independently. Then the residual noise is eliminated by PCA transform and low pass filtering of two high correlation spectrums. Experiment results demonstrate that the proposed algorithm can effectively reduce nonlinear striping in MODIS data.