Effects of Image Segmentation for Approximating Object Appearance Under Near Lighting

Shading analysis of an object under near lighting is not an easy task, because the direction and distance of the light source vary over the surface of the object. Observing a small area on the surface, however, techniques assuming far lighting are applicable, because variations of the direction and distance are small in the area. In this paper, we present two contributions to image segmentation for approximating object's appearance under near light sources. First, we experimentally evaluate the accuracy of approximations using rectangular segmentation for images of objects under near light sources, and confirm the effects of image segmentation itself. Second, we propose a novel segmentation method for approximating images under near light sources. Our proposed method plans appropriate segmentations in terms of approximation accuracy, considering properties of objects and variable illumination conditions.

• Type: Conference
• Year: 2006
• Where: ACCV
• Authors: Takahiro Okabe, Yoichi Sato
• Keywords: ACCV 2006, Computer Vision, Image Segmentation, Light Sources, Rectangular Segmentation