Shadow Editing Method based on Supervised Learning

In the process of acquisition and utilization of intangible cultural heritage, videos captured from different angles on intangible cultural heritage are always needed. But due to the insufficient illumination, there are all kinds of soft shadows on the videos. In order to remove and edit these soft shadows, we put forward to a method of shadow editing based on supervised learning. The method neither assumes the existence of umbra special model nor a unified model to deal with the whole piece of shadow, but provides the user with the complete control who are allowed to choose the area which needs modifying. We also can handle some complex problems that those previous technologies can't do, for example the shadows are all penumbra, as is often the case (such as the projection of the shadow of the leaves on the ground). Our technology requires user interaction of determining which area of the image needs to be modified. Then the system is initialized at once, then our model is applied. Once the shadow side is calculated, the user can interactively operate them, or use our simple interface to modify the rest of the images.