Material rendering property of illumination: relationship between illumination statistics and perception
Material rendering property of illumination: relationship between illumination statistics and perception
复制标题
光照的材质渲染属性：光照统计与感知的关系
DOI：
10.1167/jov.20.11.647
复制
发表时间：
2020
影响因子：
1.8
通讯作者：
Motoyoshi Isamu
中科院分区：
文献类型：
--
作者：
Kondo Daisuke;Fujita Hayato;Motoyoshi Isamu
最近的心理物理学研究表明，表面材料特性（如光泽度和颜色）的感知强烈依赖于照明统计或“光场”的空间结构。在本研究中，我们系统地研究了照明统计和感知表面质量之间的关系，以确定光场的特性，提高表面的材料外观，一个术语，我们称之为“材料渲染属性”的照明。在一系列的心理物理匹配实验中，9名观察员检查了21种不同自然光场（从大型数据库中选择的室外和室内场景）下颠簸物体的计算机生成图像，并调整了测试的感知表面特性之一-光泽度，介观粗糙度或凹凸锐度（每个维度7个级别）-以匹配参考物体的感知表面特性。例如，在“光泽度”实验的每次试验中，观察者在特定光场下自由观察给定镜面反射率（例如，0.5%）的参考对象，并在固定光场下调整测试对象的镜面反射率（桉树）。对于每个光场，我们计算了两类亮度子带图像统计，即矩统计（对比度，偏度和峰度）和能量相关性的子带。然后，我们计算了类别统计数据与不同镜面反射率下匹配光泽度变化之间的相关性。分析表明，感知光泽度的变化与光场的关键统计特征高度相关，包括对比度、偏度、峰度和交叉取向能量相关性。获得了类似的相关性介观粗糙度和凸点锐度。总之，结果表明，具有高空间对比度，偏度，峰度和交叉取向能量相关性（例如，明亮，稀疏和斑点特征）的光环境往往具有高的材料渲染特性。
Recent psychophysical studies show that the perception of surface material properties such as glossiness and color strongly depend on the spatial structure of illumination statistics, or" lightfield". In the present study, we systematically investigated the relationships between illumination statistics and perceived surface qualities in order to determine the characteristics of the lightfield that enhance the material appearance of surfaces–a term we dub the ‘material rendering property’of illumination. In a series of psychophysical matching experiments, 9 observers inspected computer-generated images of bumpy objects under 21 different natural lightfields (both outdoor and indoor scenes selected from a large database) and adjusted one of the test's perceived surface properties–glossiness, mesoscopic roughness, or bump sharpness (7 levels for each dimension)–to match the perceived surface property of the reference object. For example, on each trial in the'glossiness' experiment, the observers freely viewed the reference object of a given specular reflectance (eg, 0.5%) under a particular lightfield and adjusted the specular reflectance of the test object under a fixed lightfield (Eucalyptus). For each lightfield, we computed two classes luminance subband image statistics, namely moment statistics (contrast, skewness, and kurtosis) and energy correlations across subbands. We then calculated correlations between class statistics on the one hand and variations in matched glossiness across different specular reflectances on the other. The analysis showed that variations in perceived glossiness were highly correlated with key statistical features of the lightfield including contrast, skewness, kurtosis, and cross-orientation energy correlation. Similar correlations were obtained for mesoscopic roughness and bump sharpness. Together, results suggest that light environments with high spatial contrast, skewness, kurtosis, and cross-orientation energy correlations (eg, bright, sparse, and spotty features) tend to have a high material rendering property.