加权多尺度复合窗指纹方向场建立及分级平滑

Combining multi-scale composite windows with hierarchical smoothing method for fingerprint orientation field computation

为了有效重建低质量指纹图像的脊线信息及真实的奇异点位置,提出了一种加权多尺度复合窗的指纹方向场（orientation field,OF）建立及其分级平滑算法.方法包括加权多尺度复合窗的指纹OF建立和后续的分级平滑两步.a.建立一系列不同尺度复合窗下的OF信息;通过平方梯度一致性指标确定各个尺度在整个系列中的权重;通过权重将一系列OF信息整合得到一个较为准确的粗OF.b.将粗OF量化为两种方向模式,消去孤立的方向模式得到滤波掩膜,根据滤波掩膜对OF进行二方向模式滤波;将得到的OF量化为三种方向模式,消去孤立块得到掩膜,根据掩膜对OF进行三方向模式滤波;最后得到准确且平滑的OF.对提出算法和同类算法进行实验比较,实验结果表明：提出算法能有效重建低质量指纹图像的OF,修复错误的脊线结构并保证真实的奇异点位置不偏移.

In order to effectively construct the ridge structure and the authentic singularity in poor quality fingerprint images, a method,combining a weighted multi-scale composite window with a hierarchical smoothing strategy was proposed.This method mainly contains two procedures,which are the approximate OF estimation and the hierarchical OF smoothing.a. A series of OFs were established under multiple scales of composite windows by using gradient based method and a coarse OF was estimated by using the weight of each scale determined by a squared gradient consistency.b. The OF was first quantized into two digitized orientation zones and the two-orientation-zone filtering was adapted to the OF blocks based on a filtering mask obtained after eliminating the isolated blocks. In the end a similar three-digitized-orientation-zone filtering were performed to obtain an accurate and smooth OF. To validate the performance,the proposed method was applied to OF computation using the FVC2004 databases.The experiment results show that the proposed method is superior to the state-of-the-art methods according to their performance in OF computation. The proposed method can correct spurious ridge structure and meanwhile avoid singularity deviation compared with the previous works.