Structure-guided unidirectional variation de-striping in the infrared bands of MODIS and hyperspectral images

•A novel de-striping method for MODIS and hyperspectral infrared bands is proposed.•A novel stripe detecting method that can distinguish between texture and stripes is proposed.•Spatial information extracted using the stripe detecting method is utilized to construct spatially weighted parameters.•All parameters have the default setting, which greatly improves the practicability of our method.

Images taken using moderate resolution imaging spectroradiometer (MODIS) and hyperspectral imaging systems, especially in their infrared bands, usually lead to undesired stripe noises, which seriously affect the image quality. A variational de-striping model has been proven to have good performance, but knowing how to detect stripes effectively, especially to distinguish them from edges/textures, is still challenging. In this paper, a structure-guided unidirectional variational (SGUV) model that considers the structure of stripes is proposed. Because of the use of structure information, which textures and edges do not have, the proposed algorithm can effectively distinguish stripes from image textures and almost does not blur details while removing stripes. Comparative experiments based on real stripe images demonstrated that the proposed method provides optimal qualitative and quantitative results.