A new destriping method combining moment matching and interpolation in infrared images

Abstract

Stripe noise is very common in uncooled infrared imaging systems and often severely degrades the image quality. Based on the analysis of existing methods, a new destriping method combining moment matching and interpolation was proposed. The method mainly consisted of three steps: Firstly, moment matching was used to eliminate most stripes. Then an adaptive algorithm was used to classify the left stripes into three types: one-pixel-width stripes, two-pixel-width stripes and local stripes. Finally, these left stripes were identified and removed with weighted multi-neighborhood interpolation. The algorithm was tested with IR images heavily polluted by stripe noises, and the result indicated that it could efficiently remove stripe noise and enhance the image quality significantly better than the single method discussed in literature.