Criação de materiais compostos por BRDFs tabulares

• Ano de defesa: 2021
• Autor(a) principal: Nunes, Mislene da Silva
• Orientador(a): Carvalho, Beatriz Trinchão Andrade de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Programa de Pós-Graduação: Pós-Graduação em Ciência da Computação
• Palavras-chave em Português: Computação; Materiais compostos; Multidimensional Scaling; Bidirectional Reflectance Distribution Function (BRDF); BRDFs; Materiais; Redução de dimensionalidade; Agrupamento
• Palavras-chave em Inglês: Materials; Dimensionality reduction; Clustering
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: https://ri.ufs.br/jspui/handle/riufs/15038

In the computational representation of what is seen in the real world, appearance modeling seeks to represent how the materials reflect light in a certain direction under different lighting settings. This modeling is performed through reflectance functions. In this work, we focus on the Bidirectional Reflectance Distribution Function - BRDF, which describes the reflectance at a point on the surface through the quotient between reflected radiance and incoming irradiance on that point. This function can be represented computationally in different ways, in which stand-out tabular samples, analytical models, and linear combinations of a database of pre-existing BRDFs. As they are obtained from measurements, BRDFs represented by tabular samples present a high degree of realism at the cost of a more time-consuming acquisition process and high storage space. Aiming at creating novel and realistic-looking materials, this work proposes a pipeline to generate new materials from a tabular BRDF database. To this end, the database is preprocessed so that the main relevant reflectance features are maintained. Then, the preprocessed BRDFs are clustered in order to obtain clusters with similar reflectance features. From the selection of BRDFs from one or more clusters of interest, we propose an approach to creating novel materials which present reflectance features from the clusters of interest. This approach combines a dimensionality reduction method with a clustering algorithm. Thus, new materials were created using the proposed pipeline with different linear (Multidimensional Scaling - MDS) and nonlinear (Isometric Feature Mapping - ISOMAP and Locally Linear Embedding - LLE) dimensionality reduction methods combined with the k-means clustering algorithm. An analysis of the smoothness of the appearance of sequences of newly created materials was performed using Root Mean Square Error - RMSE.