| 其他摘要 | Inverse rendering, which recovers lighting, reflectance, and textures from photographs in order to re-render high realistic images with recovered result, is an emerging technique in computer graphics in recent years, and its application in virtual reality and augmented reality has a very good prospect. Focusing on inverse rendering, we have conducted some research on reflectance recovery for textured objects, material recovery under complex illumination conditions, and inverse global rendering from a panorama, and the major contributions include the following three aspects. 1) Reflectance recovery for textured objects For a textured object, it is very difficult to recover material and texture, since every point on the surface may have different reflectance properties. But if we category the materials into different classes we may design corresponding efficient algorithm for each class. For an object with uniform specular reflectance properties, we propose a method in recovering the global specular reflectance properties and the diffuse albedo map. The diffuse albedo map is used like a texture, to produce the fine color detail on the surface. 2) Material recovery under complex illumination conditions Generally speaking, it is very difficult to model the lighting of an object in a complex environment with a general lighting model due to the effect of indirect illuminations. Complex illumination provides more information about object reflectance, and makes it possible to recover material with less number of images. Representing complex illumination with image-based lighting, we can recover the material of a homogeneous object from a single high dynamic range image. Inter-reflection between different parts of the object is taken into account, and nice result may be obtained. Translucent and transparent materials can be also processed in such case in 3) Inverse global rendering from a panorama An environment globally illuminated is the most complicated case for inverse rendering. The environment may be complex, where objects with all kinds of materials may be in existence, and the illumination may be very comprehensive. In such case, it is difficult to give a definitive answer on how many photographs are adequate for the calculation of reverse rendering. In the present work, we propose a scheme to allow the recovery of the reflectance properties of all the objects in a scene from a single panorama, where diffuse, specular, isotropic, anisotropic, and textured surfaces may appear. Our methods can recognize textured surfaces and recover the textures from a single image. At the same time, the effect of highlights and shadows in textures may be largely reduced in the process of recovery. With the materials and textures recovered, synthetic images may be re-rendered for an augmented environment, with the illumination and the geometry of the scene changed. |
修改评论