雑音を伴う連続濃度図形から曲線を識別する一方法

概要

論文の詳細を見る
A method for line extraction in noisy photographs is presented. The method consists of the following four steps; (I) As a model of input patterns, it is assumed that input pattern=lines of uniform density + homogeneous Gaussian noise. Then the state of each sample point in the input pattern......"on a line" or "not"......is estimated by the likelihood ratio estimated from the density distribution in the neighborhood of the point. (II) Two sample points decided as "on a line" are connected if they are adjacent, thus giving connected components. (III) In each connected components,a line being most likely "on a line" is determined. The likelihood is again used here. (IV) The results are corrected according to given rules, if necessary. Three examples of applications, (i) line extraction 1,artificial noisy figures (two lines of uniform density+normal random numbers), (ii) edge detection of rib images in chest X-ray films, (iii) artificial binary patterns (circle + uniformly distributed noise) are shown with considerably successful results.
一般社団法人情報処理学会の論文
1970-07-15