基于可分解的卷积核和快速傅立叶变换FFT的指纹图像增强算法

Fingerprint Image Enhancement Based on Decomposable Kernels and Fast Fourier Transforms

指纹识别系统在图像增强阶段涉及很多卷积运算,占用大量的计算时间。在不利用专门的DSP处理器进行指纹图像增强处理情况下,常规的卷积运算计算时间非常长。在指纹图像增强过程中,卷积运算主要集中在方向场估计和高波滤波两个阶段。为此本文提出了一种指纹图像增强算法,利用可分解的卷积核和快速傅立叶变换来替换常规的卷积运算,可减少算法的时间复杂度,快速实现指纹图像增强。

In a fingerprint verification system, enhancing fingerprint images needs lots of convolutions which take plenty of computation. Without help of any specialized digital signal processor to tackle this task, it takes a very long time for generic convolutions. There are two stages in the fingerprint image enhancement involved with heavy convolution. In this article, instead of classic methods, we use decomposable kernels and fast Fourier transforms to realize convolutions in the stages of estimating orientation field and Gabor filtering for fingerprint image enhancement, which enable us to reduce the time complexity and achieve a faster computation.