Desriping Hyperion Imagery Using Spline Interpolation

This paper presents an alternative approach for removing striping patterns of Hyperion hyperspectral imagery. In this study, a cubic spline-based curve-fitting algorithm was used to estimate gray values of striping pixels in an image. The idea is to treat each line of a Hyperion image as a piece-wise spline curve in each spectral band. After identifying pixels affected by the striping noises, the noise-free pixels are collected and used as samples to construct a cubic Hermite spline curve that passes all of the control points (samples). The original gray values of striping covered pixels can then be approximated from the spline curve. Applying this procedure to an image line by line or column by column, a striping-free image can be reconstructed accordingly.