簡単なデータ表現を用いた両眼視に基づく距離情報の抽出

概要

論文の詳細を見る
Depth information is important to the three-dimensional recognition of a scene by computers. In this paper, a depth recognition method is described which is based on the binocular stereoscopic vision. Extraction the depth information from steroscopic images requires algorithmic matching of the left and right images, the so-called correspondence problem. The proposed algorithm is based on finding boundaries of regions corresponding each other. The regions are defined as the parts with certain gray-levels in each of the images. Thus, the given images are transformed into 2-valued images. This enables the proposed simple algorithm (region-method) to get correspondence of objects in both images. The region-method is applied to two examples and compared with a conventional correlation-method.
一般社団法人情報処理学会の論文
1977-03-15