曲面画像平面化処理における立体形状推定誤差に対する画像情報認識の精度検証

Abstract

Image information to be obtained from the target in real world may be distorted because they are on the curved surface. Existing techniques of image processing are designed for planar images, and are not suitable for distorted images. To apply existing techniques for distorted images, planation methods are desired. These methods need 3D structure information to be stretched into plain surface, but estimated 3D information obtained by structure measurement tools such as stereo vision or SfM(structure from motion) are represented with low resolution. In order to obtain planared image using low resolution 3D information, we proposed grid expansion technique with two sampled grids of different phases. Experimental results showed the effectiveness of our proposed method.