An original multi-sensor approach to scale-based image analysis for aerial and satellite images

This paper presents an original scale-based image analysis scheme dedicated to the automatic low-level interpretation of aerial and satellite images. It is based on an entropic scale detector. Since this detector is highly robust, the scheme we propose can efficiently merge images from sensors of various resolutions, and thereby widen the usual range of scales that can be processed. Hence, it is adapted to the analysis of remote sensing scenes, that contain objects of a wide range of sizes. This original method efficiently merges the information of various scales, while being robust and fully automatic. Eventually, it detects every object with a minimum amount of pixels, and is thereby fast.