BTF Prediction Model using Unsupervised Learning

The impressions evoked by textures are called affective textures, and are considered to be important in evaluating and judging the quality of an object. And, technologies for understanding and controlling sensory textures are needed in product design. In this study, we propose a BTF prediction method using DNN as a first attempt to generate textures based on affective texture recognition. The method uses a series of continuously varying viewpoint angles of a texture image as the input signal. This method enables the generation of texture images with continuously changing angles. We tested the validity of the proposed method by using textile, wood and paper. The results show that the proposed method is effective for predicting diffuse reflection optical properties and irregular and regular patterns.