抄録
This paper describes the process of recovering a 3D shape from stereo image. Two CCD cameras are set up and a parametric curve is on an epipolar plane. A superconic function is used as the parametric curve. This curve causes disparity among the right image and left image. The visible part of the curve shows the object's profile which is cut by an epipolar plane. The estimation function is defined using the disparity including parameters. Parameters of the curve are controlled to optimize the estimation function. The estimation function is defined in the following steps : 1st : the visible part of the curve is calculated. 2nd : the visible part is projected to the right and left images. 3rd : the length of the projected curve is calculated and it is compared with the two images. 4th : the length and difference of the two images' intensities are used as the estimation function.
Parameters of the curve are controlled to optimize the estimation function using the polytope method. This algorithm is tested with experimental image data. The results are similar to the DP matching method which resolves the correspondence problem.
