基于D-S证据理论的指纹图像分割方法

Fingerprint Image Segmentation Based on D-S Evidence Theory

在对D S证据理论进行简单分析之后 ,针对指纹图像提出了一种基于D S证据理论的图像分割算法 .该方法使用方向性和对比度两个信息分别作为两个分类器的特征 ,并利用模糊规则求出各分类器的基本概率分配函数 ,最后利用D S证据理论的合成法则将两个分类器的结果进行融合判决 .该方法具有较高的稳健性和精确度 。

Segmentation is an important step of image preprocessing. Effective segmentation can not only reduce the time of subsequent processing, but also improve the reliability of feature extraction considerably. After simply analyzing some essential knowledge of D S evidence theory, authors suggest a novel segmentation method for fingerprint image based on D S evidence theory. In this method, two threshold segmentation is used to segment the non ridge areas of the images, and direction and contrast are respectively selected as features for two separate classifiers based on the characteristic of the fingerprint images. The feature which is extracted from the image areas is usually uncertain. In order to describe and process this uncertainty, the method uses variable’s membership functions to describe the reliability of proposition and basic probability assignment functions of the two classifiers are acquired. At last, the combination discipline of D S evidence theory is used to determine the final result of all the classifiers. Thus the clear fingerprint areas, noisy fingerprint areas and fussy areas are achieved. Experimental results verify the feasibility and robustness of this algorithm.