The effects of lighting direction and elevation on judgements of shape-from-shading.
The effects of lighting direction and elevation on judgements of shape-from-shading.
复制标题
照明方向和高度对阴影形状判断的影响。
DOI：
10.1167/12.9.234
复制
发表时间：
2012
影响因子：
1.8
通讯作者：
Mazzilli G
中科院分区：
文献类型：
--
作者：
Mazzilli G
照明方向的知识是至关重要的恢复形状从阴影。许多研究表明，当真实的光照方向未知或不明确时，我们会采用从上方照射的假设。其他人已经表明，感知到的照明方向取决于阴影图案本身中的取向结构。许多研究已经测试了照明方位角变化对阴影恢复形状感知的影响。我们测试了改变光源的倾斜度（高度）的效果。我们生成了三个不同表面（各向异性程度不同）的图像，这些表面用聚光灯以6个倾斜角（0度，正面到75度，倾斜）照射，方位角不变。这些表面生成的Gabor噪声纹理的表面高度确定的“灰度”值的每个纹理。我们在两个方向上测试了图像：直立（表面从右侧照亮）或旋转（从上方照亮）。不同的decompression会影响最终阴影图案的对比度，因此我们还在归一化RMS对比度后测试了图像。观察者设置探针盘的倾斜度和斜度以匹配感知到的3D表面取向。我们在每个图像的九个位置进行了观察。我们发现，最各向异性表面的倾斜设置取决于图像方向，而各向同性表面的倾斜是由从上面的照明之前。倾斜设置倾向于与局部亮度/对比度成比例地增加，并且标准化RMS对比度大大减小了倾斜设置的范围。总之，我们发现，在各向同性的阴影图案的形状判断是更受影响的照明从上面的先验比那些各向异性的表面和倾斜的判断取决于当地的亮度/对比度，并受光源的decompensation的变化，当这影响的阴影图案的对比度。
Knowledge of the lighting direction is crucial in recovering shape-from-shading. Many studies have shown that we adopt a lighting-from-above assumption when the true lighting direction is unknown or ambiguous. Others have shown that perceived lighting direction depends on the orientation structure in the shading pattern itself. Many studies have tested the effects of lighting azimuth variations on the perception of shape-from-shading. We test the effects of varying the declination (elevation) of the light source. We generated images of three different surfaces (differing in their degree of anisotropy) illuminated with spotlights at 6 angles of declination (0, frontal to 75 degrees, oblique) with constant azimuth. These surfaces were generated from Gabor noise textures with surface heights determined by the ‘grey’values of each texture. We tested images at two orientations: upright (with the surface lit from the right), or rotated (lit from above). Varying declination affected the contrast of the resulting shading patterns so we also tested images after normalising their RMS contrasts. Observers set the tilt and slant of a probe disk to match the perceived 3D surface orientation. We obtained observations at nine locations per image. We found that tilt setting for the most anisotropic surface depended on image orientation whereas for the isotropic surfaces tilt was determined by the lighting-from-above prior. Slant settings tended to increase in proportion to local luminance/contrast, and normalising RMS contrast greatly reduced the range of slant settings. In conclusion we found that shape judgments in isotropic shading patterns are more affected by the lighting-from-above-prior than those for anisotropic surfaces and that slant judgments depend on local luminance/contrast and are affected by changes in the declination of the light source; when this affects the contrast of the shading pattern.