基于灰度特性的指纹图像预处理的研究

【作者】 王剑峰；

【导师】 曾巧明；

【作者基本信息】 中南大学 ， 计算机应用与技术， 2008， 硕士

【摘要】 在自动指纹识别系统中,由于各种原因使得采集到的指纹图像有10%左右是低质量指纹图像,这对细节特征的提取造成了困难,同时也影响了特征提取的准确性。为保证指纹识别算法对指纹质量的鲁棒性,需对采集的指纹图像进行预处理处理以提高指纹图像的质量。论文对指纹图像的预处理进行了研究,其主要工作如下:首先考虑对指纹的分割处理,在进过分析了基于灰度统计的分割算法后,将自适应均值和标准差的比值作为阀值分割的标准,并在分析了残留区域与有效区域的关系后,引入了轮廓跟踪来进行二次分割;其次着重从方向图计算和滤波处理两个方面对指纹图像的增强进行研究,在方向图的计算时,通过对几种基于直接量化法的方向图的计算方法进行分析和比较,给出了改进的算法,然后在方向图的基础上设计了一套方向模板对指纹图像进行滤波;最后对多种细化的算法进行比较、分析,结合快速细化算法和OPTA算法对二值化图像进行细化处理。实验结果表明,论文中的算法能够有效的将指纹的有效区域分割出来,在噪声的影响下可以更准确的求取图像的方向图,通过增强处理可以有效地增强指纹图像中沿着脊线方向上的信息,减弱与脊线方向垂直方向上的信息。细化后的骨架接近纹线,更有利于特征点的提取。更多还原

【Abstract】 In practice, due to kinds of cause, a 10 percentage of acquired fingerprint images are of poor quality by fingerprint image collecting in the system of automated finger identification. It’s hard to extract the minutia from the poor quality fingerprint, and also the accuracy of minutia is affected. In order to ensure that the performance of the quality of input fingerprint image, a series of processes are needed to enhance the quality of fingerprint image.The paper discussed the fingerprint image preprocessing. The main research are as following: In the step of the segmentation, after analysis the method of segmentation based on gray-scale statistic, the segmentation use the ratio of the adaptive average value and standard deviation as the threshold of standard and then import the boundary of tracking after discussing the relation of remain region and effective region. And then the fingerprint enhancement is discussed from the side of direction computation and process of filtering, after comparing some methods of directional image computation, an improved method is proposed based on joint cutting method, designing a set of directional template to filter the fingerprint image. The set of methods can enhance the quality of fingerprint image effectively. After comparing and analyzing some methods of binarization and thinning, locally adaptive thresholding is chosen to binarize the enhanced image, and combining the quick thinning algorithm and improved OPTA thinning for the process of thinning.Experiments show that the methods of preprocessing in this paper have good results, which have effectively divided foreground form the fingerprint image and computed direction of fingerprint under the circumstance of voice. And after enhancement, the information on the ridgeline is effectively enhanced, while the vertical is weakened. The skeleton after thinning is convenient to extracting character form it.更多还原