基于学习的NSCT的指纹图像超分辨率重建  

Fingerprint Image Super-Resolution Reconstruction Based on NSCT Learning

作　　者：吴巧玲[1] 倪林[1] 何德龙[1] 刘权[1] 

机构地区：[1]中国科学技术大学电子工程与信息科学系,合肥230027

出　　处：《数据采集与处理》2012年第2期173-178,共6页Journal of Data Acquisition and Processing

基　　金：国家自然科学基金(61172157)资助项目

摘　　要：非下采样Contourlet变换(Nonsubsampled contourlet transform,NSCT)采用非抽样金字塔结构和非抽样方向滤波器组构成,具有Contourlet变换所不具备的平移不变性、较高冗余度等优良特性,而且能够克服伪吉布斯现象。图像经过非下采样Contourlet变换后分解成多尺度、多方向的细节信息,这些细节信息代表了图像不同频带不同方向的特征,这就简化了系数之间的关系。基于学习的超分辨率重建算法具有整体的预测性,将非下采样Contourlet变换和基于学习的算法相结合,在一定程度上提高训练精度。针对指纹图像的实验证明该算法具有良好的性能,重建的图像纹理性细节信息较好,基本保持了原指纹图像的特征点,更接近于原始的高分辨率图像。The nonsubsampled contourlet transform （NSCT） is based on a nonsubsampled pyramid structure and nonsubsampled directional filter banks. NSCT is shift-invariant,of high redundancy compared with the contourlet transform,and it can overcome pseudo-Gibbs phe- nomenon. An image can be decomposed into multiscale and multidirectional details. These details represent image characteristics in different directions and different frequency bands, which simplifys the relationship between coefficients. Learning-based super-resolution reconstruction algorithm has the predictability of a whole image, and learning-based method combined with NSCT can improve training accuracy. Experiments with fingerprint images show that the algorithm has good performance. Furthermore, the reconstructed images can maintain the feature points possessed of the high resolution images and they are closed to the original image with good texture details.

关 键 词：图像超分辨率 指纹 非下采样CONTOURLET变换 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]