局所並列演算による濃淡図形の縁の線画化 (<小特集>パターン情報処理)

概要

論文の詳細を見る
Edge extraction is one of the most important and fundamental technique in picture processing. A number of methods, based on different concepts, have been proposed. But simple methods depend too heavily on a priori knowledge of the input and the all-round methods depend too heavily on a priori knowledge of the input and the all-round methods are often unreal. This paper presents a method which extracts edges by a finite sequence of local parallel operations. The main concept of extracting edge points is to select a point whose edge value is maximum on two or more of the four directions crossing at that point. This technique is very sensitive to noise, especially in rather flat area. To remove this defect, the noise level is estimated using the least square linear approximation and used to modify the edge value obtained by a simple method. The characteristics of the method are studied using the natural patterns and the artificial one.
一般社団法人情報処理学会の論文
1976-07-15