3D Material Style Extraction from a Single Image

Material design is an important component of 3D scene modeling,which however costs a lot of time and labors.In this paper,we present a new method for automatically extracting the material style from a sample image for 3D scenes.Given a 3D scene and a sample image,we first obtain some candidate materials using inverse rendering techniques under a set of pre-determined views.Then,we define an objective function to estimate the difference of the material style between the sample image and the synthesized images rendered with candidate materials.This objective function takes into account the difference of the mean,the variance and the histogram of the chromatic distribution.Finally,we select best materials as candidate materials by optimizing the objective function using the Tabu Search method.To validate the effectiveness of our method,we compare our results with those obtained with user assistances to demonstrate our method.