A Fast Shadow Detection And Removal Algorithm For High Resolution Satellite Images

In recent years, as more and more high-resolution satellite images being used in the remote-sensing field, remote-sensing technologies start to play an important role in many areas, such as disaster monitoring, earthquake rescue and ground changes monitoring. However as the resolution of the satellite images increasing, the influence of the shadows inherent from the images becomes more and more obvious. Whenever an area in the remote-sensing images is blocked by some object from the light source, a shadow it becomes, which usually means losing information from images. And this makes certain image processing and analysis task becomes hard. In fact, there is still some information reserved in shadow areas, which is critical to increase the value of the high resolution image data.The shadow processing technique includes two parts which are shadow detection and shadow removing. The shadow detection technique is already in use of some computer vision applications. Due to the spreading use of digital images these years, more and more attentions has been paid to this technology from different field, such video processing and traffic controlling. Basically, there are two approaches to detect shadows from images, which are image-based methods and object-based method. During the two, the latter one is rarely used in practice, because it relies on some prior knowledge or the digital model of the scene in the image. As for shadow removing techniques, most existed methods take the digital image processing approach, which make compensation to those pixel values in shadow area.Through analyzing the image data in various color space, this paper propose a new shadow detection and removing algorithm based on the color space transformation which is highly practical. Comparing with those traditional methods, our algorithm features improved detection precisions, high flexibility that could be used in various circumstances. In the shadow removing part, our method can quickly restore the area information that is covered shadows through statistic approach.