Material Appearance SVBRDF Data Acquisition And Modeling Methods

In recent years,the application of photorealistic rendering in life and production has become more and more extensive.Among them,the material is an important part of the photorealistic rendering,and the appearance of the material in the real world is displayed to people.In the natural world,opaque materials occupy a large proportion.SVBRDF models are commonly used in computer graphics to express such materials.They generally show irregularities in texture and reflectivity.At present,in the industry,experienced designers generally draw and adjust manually.However,restoring the reflection characteristics of the material through the acquisition method can greatly improve the production efficiency.Therefore,it is of great significance to study the SVBRDF acquisition and modeling method.Aiming at the collection and modeling of SVBRDF,this thesis conducts the following research:1)Designing and builting a set of hemisphere acquisition system for SVBRDF data collection.By controlling 9 industrial cameras and 196 light sources uniformly distributed on the hemisphere,material images can be collected densely from different height angles and azimuth angles.Through mechanical and optical optimization design,parallel collection,multi-threaded programming and other means,the device has a higher collection accuracy and faster collection speed.2)Estimating SVBRDF using photometric stereo and iterative refinement clustering.For the SVBRDF data collected by the hemisphere device,the photometric stereo method was first used to solve the material normal map,and then the iterative refinement clustering algorithm was used to optimize the diffuse reflection,specular reflection and roughness maps,and the image selfsimilarity prior was used.The common fitting of pixels with the same reflection properties reduces the risk of estimation failure caused by abnormal data collected by a few pixels in the pixel-by-pixel solution.3)Acquising and estimating SVBRDF using ambient and flash images.Aiming at the problem that the textures are prone to artifacts when estimating smooth materials in the existing methods,use a portable device to take a flash image and an ambient image,and re-estimate the diffuse,roughness and specular maps to make the roughness and specular map more close to physical reality,the rendered image is closer to the real image.