Single Image Light Positioning Based On Render

Image understanding is a very important branch of Computer Vision. If we want to understand the image more accurately, the determining of light position is a very important part which can't be ignored. We can recover the information of image by adding three-dimensional objects with the appropriate light in rendering software.Therefore, this article proposes Single Image Point Light Positioning System based on Render. By understanding the three-dimensional model out from the source image, and then rendering at different light position in rendering software, we can get the location of the input source image ultimately estimates.It is a reverse progress to get the information from an image like getting solution. Since we introduced image rendering, we can determine an image with the three-dimensional model and light source information. So the progress to get the light source position becomes a forward solution testing progress. This article proposes a method that we render an image at every position in space, and then compare the rendering image with the source image, if they nearly the same, we can say the position is the best estimate of the light source position. And this method makes the light source localization becomes more easy to understand and have higher reliability.This paper analyzes the necessity and feasibility of the source localization progress in image understanding. We described a light source positioning system based on a general algorithm and analyzed the limitations and shortcomings of this system. This article introduced the progress of the light source positioning system based on rendering system, related evaluation of image rendering and gray theory. Then we analysis limitations encountered in the real image experiment. Finally, we designed an algorithm which greatly improves the light source positioning system.