图象分割的自适应模糊阈值法

Histogram Adaptive Thresholding Using Index of Fuzziness

该文研究了海面舰船图象的模糊阈值分割问题 .首先介绍了模糊阈值分割的基本原理 ,在讨论了隶属函数的分布特性及其窗宽对阈值选取的影响后 ,提出了一种在预先给定隶属函数的情况下 ,利用目标 -背景对比度自动选取窗宽的方法 ,并给出了根据目标与摄像机间的相对距离估计目标 -背景对比度的算法 .其应用于智能电视跟踪系统 ,对不同距离、不同对比度的海面舰船图象进行了阈值分割实验 ,结果表明该方法具有较强的场景适应能力 .

A novel approach for adaptive threshold selection from gray level histogram using index of fuzziness is presented. The principle of threshold selection through minimizing index of fuzziness is described first, and then how to choose the membership function and window size is illustrated. Based on the object background contrast computed from the distance between the object and the camera, a new method for the automatic choice of window size of a predetermined membership function is provided. Also an algorithm to obtain object background contrast from the distance between the naval vessel and the camera is suggested. The experiment is conducted over three images of a target warship with different distance between the object and the camera and with different contrast. The results show that the approach proposed herein has better performance on adapting the scenario variance.