基于多通道Gabor滤波的指纹图像二值化方法

为解决传统的直接利用图像全局或局部灰度特性的指纹图像二值化方法的不足,提出一种多通道Gabor滤波的指纹图像二值化方法。首先构造多通道Gabor滤波器组对规格化后的指纹图像提取不同方向纹线对应的主要频率成分。将获得的多通道的滤波响应图线性叠加形成单幅Gabor滤波叠加响应图。然后对其再次采用以上滤波器组滤波,分别求取各位置的最大Gabor响应构成最终的指纹滤波能量响应图。最后分块并求取块内阈值实现二值化。实验结果表明,该方法较基于灰度特性的方法鲁棒性更好,特别对存在粘连、噪声和断纹、孔洞等现象的指纹图像有较好的处理效果。

A Method for Fingerprint Image Binaryzation Based on Multichannel Gabor Filters

In order to overcome the drawbacks of traditional fingerprint image binaryzation methods that utilize global or local image gray characters directly, a new method based on multichannel Gabor filters is proposed. Firstly multichannel Gabor filter banks were constructed to segregate the texture frequency domain bases of the normalized image from different directions. Secondly the obtained filtering response diagrams were processed with linear superposition in order to form a single stack response image of Gabor filtering, and then we filtered the image with the same filter banks and evaluated the maximal Gabor responses of each slot to constitute the final energy response diagrams of fingerprint filtering. At last piecemeal the diagrams and cumulate the thresholds to realize binaryzation. The experimental results show that the method has stronger robustness than any other methods which use gray characters, especially have better results to the images that have the phenomena of adhesions, noises, creases and holes.