MSER based shadow detection in high resolution remote sensing image

The shadows are regarded as obstacles in remote sensing image analysis. With high-resolution remote sensing images developed, especially in urban area, shadow detection plays a much more important role in many applications. This paper presents a novel vision-based shadow detection method. The shadow areas are usually much darker than non-shadow areas visually in high resolution images. In our method, we extracted MSER (Maximally Stable Extremal Regions) in the image. Then these regions are classified into shadow-areas and non-shadow areas. Our method is experimentally verified by applying it to Quickbird images. The experimental result shows that it can effectively extract shadow areas in high resolution remote sensing images.