中文摘要：

获得准确的方向图是自动指纹识别系统（AFIS）预处理的重要一步。基于梯度的方向图估计算法的是常用的一种有效算法，针对其不足之处，对其进行了改进。首先采用4个方向的梯度分量来计算每点的梯度和方向；采用自适应的平滑模板来适应不同质量指纹的方向图平滑需要；综合不同平滑程度的方向图增强去噪能力；同时对中心点（corepoint）、三角点（deltapoint）附近采用不同的平滑方法，使这些区域纹线变化细节得到保持。实验结果表明改进后的算法可以获得更准确的方向图，去噪能力增强，可以适应不同质量的指纹图像。

英文摘要：

It is critical to estimate the orientation field in Automatic Fingerprint Identification System（AFIS）. Gradient-based method is widely used to estimate orientation field. An improved method is proposed. First, four directional gradient components are used to compute the gradient and direction of each pixels. Then a template with adaptive size and coefficients is used to smooth the orientation field of fingerprints with different quality, and two smooth orientation fields of different scale are synthesized to reduce noise further while preserving the details. A different smoothing method is used in the core or delta areas to preserve the orientation details in these areas. The experiment results show the accuracy of the improved method and the improvement in removing noise.