Research On Materials Measured Data Based Reflectance Modeling Algorithm

Photorealistic rendering technology is an important issue in computer graphics, and realistic light effect is main objective of photorealistic rendering technology. So the material reflectance attribute is key problem.Currently, people have proposed many methods that based on mathematical analytical model or based on material measurement data. These two kinds of methods have different pros and cons. On one hand, the methods based on mathematical analytical model are simple in data processing, but the rendering results are not very fine, so it is difficult to meet highly realistic rendering requirements. On the other hand, although the methods based on material measured data can show fine rendering results, they always need acquisition equipment, huge data storage, and time-consuming complicated data processing technology.Therefore, researching and application of measure-based material is more and more common, and there are some methods that use BRDF model to fitting the measurement data. However, the performance of BRDF analytical models to fit measured BRDFs is better than measured BTFs, because BTF accommodates self-shadowing, inter-reflection, masking effects of a complex material.This thesis implements a method that uses BRDF analytical models to fit measured BRDFs data and measured BTFs data. The main work of this thesis is divided into two parts:Firstly, we implement measured BRDFs reflectance modeling. We use different BRDF analytical models and different error metrics to fit MERL BRDF database. Then, we compare the performance of analytical models and error metrics in terms of their ability to fit measured BRDFs and find best analytical model and error metric for measured BTFs reflectance modeling.Secondly, we implement measured BTFs reflectance modeling. We decompose measured BTFs into meso-scale structure and micro-structure, and implement a fitting method using BRDF models based on shadow removal for the micro-structure. First of all, we extract the meso-scale geometry information from measured BTFs. Next, we use the geometry information and lights information to compute the shadow information and remove the pixels in shadow area for each picture. Then, we take each pixel as a BRDF data and use BRDF models to fit it. Because directly applying the algorithm to each pixel of a BTF would require a prohibitively long computation, we do a k-means clustering over all pixels. In the end, we use the BRDF models and geometry information to render the new effect of the object under new viewpoint and illumination. Compared to the current reflectance modeling method of the measured BTFs with masking, shading data, our method is more close to the measured BTFs so as to have more realistic visual effects.