Tik-111.500 Seminar on Computer Graphics Spring 2003: Image-based Methods

Most graphics applications today require more or less photorealistic scenes to be rendered. This can be quite easily accomplished by rendering 3D models with increasing complexity. However, despite the fact that geometric modeling can nowadays be hardware accelerated, photorealistic results need a lot of computing and thus interactivity is usually sacrificed. Therefore, in the last few years the usage of images in 3D graphics has been extended. Numerous modeling and rendering methods have been developed that rely on images, creating the sub-field of image-based rendering (IBR). IBR uses captured images of the environment and creates new views by resampling the images that already contain the geometric and photometric properties. One of the great appeals of IBR is the potential to alleviate the modeling bottleneck as images (photographs) can be easily acquired. One particularly promising technique is image-based rendering by warping (IBRW). It uses 3D warping equation that takes samples from the reference image(s) and maps them to destination image. Recently, there have also been proposals of IBRW algorithms that could also be hardware accelerated. WarpEngine is one example of this kind of a system.