基于改进型D-S证据理论的决策层融合滤波算法

本文针对线性和非线性混合滤波技术的不足,提出了一种基于改进型D-S证据理论的融合滤波算法.本算法首先分析了基本D-S证据理论组合准则缺陷的产生原因,从而采取消去证据间相关性、自适应地分配矛盾信息给与冲突有关的焦元等措施,改进了基本D-S证据理论的组合准则,提高了其融合性能.在此基础上,本算法又根据混合滤波器单源判别边界点误报风险大、可靠性和容错性差的弊端,增加了多子源判决准则,利用改进了的D-S证据理论的组合准则,实施融合判决.后续实验证明了本算法的有效性,为图像滤波的技术发展提供了一个新的探索途径.

Decision-Level Fusion Filtering Algorithm Based on Advanced D-S Theory of Evidence

A fusion filtering algorithm based on D-S theory of evidence is proposed to combat shortcomings of linear and nonlin-ear filtering techniques. It first analyzes how basic D-S theory of evidence can result in limitations and to what degree, and then takes measures to annul dependencies among evidences and adaptively assigns eontradictionary information to conflict-relevant pixels, thus improving the combination criteria of D-S theory of evidence as well as increasing its fusion capability. On this foundation, and for thedisadvantage in hybrid fdters of great risk of misjudging image edges, lower reliabihty, and had error tolerance, we add a multi-sub-source adjudgement criteria, and utilize modified combination criteria in D-S theory of evidenced to perform fusion adjudgement. Subse-quent experimental results show the validity of the algorithm, and thus providing a new way to improve image filtering techniques.