Hyperspectral image denoise based on curvelet transform combined with weight coefficient method.

In order to overcome deficiency in wavelet analysis, people have proposed beyond wavelets based on wavelets transform, in which curvelets transform is a kind of commonly used method for signal processing. Because of directional sensitivity of its basis function, curvelets transform suggested more superior characteristics than that of wavelets in describing singularity for curve and image edges. Because the spatial distribution of stripe noise in hyperspectral image has clear directivity, so it is more reasonable that it uses Curvelet transform for image denosing. Combining with spectrum relation, curvelet transform was used in this paper for noise elimination of the hyperspectral image. The algorithm used in this paper includes three main steps. Step 1: Weighting sums of high frequency curvelet coefficients in bands containing less noise; Step 2: Replacing the high frequency curvelet transform to obtain reconstructed images with less noises. Experiments indicated that this method can carry on the effective denoising to hyperspectral image, and retained detail information well at the same time. The denoising images obtained by this method possessed a higher peak signal-to-noise ratio and better visual effects.