A solution to histogram-equalization and other related problems by shortest path methods

被引：8

作者：

Kundu, S ^{[1
]}

机构：

[1] Louisiana State Univ, Dept Comp Sci, Baton Rouge, LA 70803 USA

来源：

PATTERN RECOGNITION | 1998年 / 31卷 / 03期

关键词：

shortest path; histogram equalization; step-function approximation;

D O I：

10.1016/S0031-3203(97)00049-6

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

Given a list C = [c(1), c(2), ... , c(N)], c(k) = f(x(k)) greater than or equal to 0 and x(1) < x(2) < ... x(N), the histogram-equalization problem consists of decomposing the list C into M less than or equal to N pieces or intervals, where each interval is a group of one or more consecutive c(k)'s, such that the sum of c(k)'s for the various intervals are as uniform as possible. We give an O(MN2) algorithm for computing an optimal decomposition by converting it to a shortest path problem in a suitable digraph and using a modified form of Floyd's algorithm for computing a shortest path. A number of other decomposition problems, which have applications in curve fitting by step-functions and image compression, can be solved by the same method. (C) 1997 Pattern Recognition Society. Published by Elsevier Science Ltd.

引用

页码：231 / 234

页数：4