A Fast Image Mosaicking Method Based on Iteratively Minimizing Cloud Coverage Areas

Abstract

There are many areas with cloud coverage in remote sensing images frequently, and they will shelter the ground objects under clouds. When images with clouds are used in mosaicking, cloud coverage areas may still exist in mosaics, which constrain the application of mosaics. Therefore, a simple and fast image mosaicking method based iteratively minimizing cloud coverage areas is proposed in this letter. In the proposed method, both cloud-free images and images with clouds are used to generate a mosaic with minimal cloud coverage areas. After the generation of initial effective mosaic polygons (EMPs) for images, a quadtree iterative procedure is presented to optimize each EMP by excluding the residual cloud coverage areas in each EMP. The procedure fully takes into account the spatial distribution of cloud coverage areas, and each area with cloud coverage is minimized by as few images as possible. The experimental results show that the mosaic generated by the proposed method can achieve a better radiometric continuity and higher efficiency than the pixel-based mosaic method. Besides, the proposed method can make the full use of the cloud-free areas in images, which can improve the utilization of images and the currency of mosaics.