基于显著特征的指纹自动定位与匹配算法

非约束图像的自动匹配问题包括定位和配准两个关键步骤,其中定位是配准的前提。指纹图像的自动匹配是一种典型的非约束图像匹配问题。由于指纹采集时按压指面的旋转、变形和扭曲,以及各种因素造成的图像模糊、残缺和特征信息丢失,使指纹图像存在较大程度的不确定性,给利用细节特征进行匹配带来了困难。该文通过分析特征匹配的一般问题模型,指出了匹配问题的实质是在两个不完全等价特征集中寻求最大同构子集的问题,并对目前发展起来的几种匹配方法的适用性做了一定的评析。结合指纹图像的自身特点,提出了一种新的基于显著特征的自动定位与匹配算法。该算法通过分析细节特征点之间的联系,在整幅图像范围内构造了3种类型的显著特征,通过这些显著特征之间的比对实现了两幅图像的快速自动定位,求出了图像平移和旋转的变换矩阵。最后通过特征点的平移和旋转变换实现了具有容忍局部误差的细节特征点的比对。这种算法具有较强的容错性,对各种噪声图像具有一定的鲁棒性,是一种自适应的快速匹配算法。理论分析与实验结果表明这种方法用于指纹图像进行细节点匹配是很有效的。

Algorithm on Fingerprint Automatic Locating and Matching Based on Significant Features

The problem of automatic matching for non-restraint images involves two key steps:the previous locating and the following matching.Fingerprint image is a typical non-restraint image that has some uncertainty because of the swivel,distortion and all kinds of noises,which make it difficult to perform matching employing the minutes.This paper states that the substance of matching is to seek the maximum isomorphic subset in two non-equivalent feature points sets through the analysis of features matching model and appraises several existing methods.According to the characteristics of fingerprint image,a new algorithm to locate and match images is presented.By constructing three types of line features in the complete image,the translational matrix and swivel matrix can be gotten.After these two kinds of transform,the authors can achieve matching according to the position of minutes within the range of a permissible error.This algorithm is an adaptive and rapid matching method with good fault tolerant property.Theoretical and experimental results show this method is also effective for noisy images.