【作者】 李继杰；

【导师】 黎蔚；

【作者基本信息】 河南科技大学 ， 计算机软件与理论， 2009， 硕士

【摘要】 手掌静脉图像是一种生物特征识别技术,在识别过程中具有活体识别、内部特征、非接触式、不可伪造性等优点,成为身份认证的研究热点。但是,手掌静脉在获取和传输过程中,通常会出现噪声附加以及图像失真等问题,为了给手掌静脉识别系统提供可靠的特征依据,需要对采集到的手掌静脉图像进行增强和降噪处理。Contourlet变换作为一种新的多尺度分析方法,不但具有多分辨率时频分析特性,而且具有良好的各向异性和方向选择性,能够用较少的系数来表示光滑的曲线,是一种“真正”的二维图像稀疏表示方法。本文围绕Contourlet变换在手掌静脉图像降噪和增强中的应用进行研究,主要工作如下:(1)详细介绍了Contourlet变换和非下采样Contourlet变换的基本原理,为后续降噪和增强算法的研究奠定了基础。(2)分析了噪声对手掌静脉图像的影响,针对传统Contourlet阈值降噪方法中存在的“过扼杀”现象,提出一种基于Contourlet变换的手掌静脉图像去噪算法。对分解系数作尺度积,使系数具有较强的相关性,根据各尺度不同方向子带的能量分布确定阈值,在近似分量和细节分量上分别采用S型函数和收缩阈值函数对系数进行处理,从而在降噪的同时避免了过度的扼杀边缘信息的现象。实验结果表明本算法去噪效果良好。(3)分析了手掌静脉图像质量分布不均的问题,针对图像进行局部直方图均衡化处理后出现的局部区域过度模糊、噪声附加等问题,采用自适应窗口的方法改进了均衡化算法;并将该算法应用到非下采样Contourlet变换增强算法中,在增强图像的同时抑制噪声;最后通过Gabor滤波器增强手掌静脉纹路信息。实验结果表明本算法增强效果比较明显。更多还原

【Abstract】 Palm vein image as a cutting-edge technology of biometric trick identification, it has the characteristics of trueprint, non-contact and unforgeability during the identification, so it becomes the research hotspot of authentication. But the problems of the noise and image error often appear in the process of the acquisition and transmission for palm vein. In order to provide the credible characteristic basis for the system of palm vein identification, it is necessary to enhance the collected images of palm vein and implement the noise reduction.Contourlet transformation as a new multi-scale analysis method, it not only has the multi-resolution and time-frequency analysis features, but also has good anisotropy and directional selectivity, and it can express the smooth curve with less coefficients, which is a true of sparse representation of two-dimensional images. This paper revolves around Contourlet transformation in the application of noise reduction and enhancement for palm vein image. The main tasks are as follows:(1) Described the basic principle of Contourlet transformation and Contourlet transformation of non-down-sampling in detail, which lay the foundation for the next study of the noise reduction and enhancement algorithms.(2) Analyze the noise of palm vein image. According to the "over kill" phenomenon of the traditional threshold denoising method, a noise reduction algorithm of palm vein image based on Contourlet transform is proposed. The shrinkage factor is used to distinct the information coefficient and noise coefficient in the approximate area. The semi-soft threshold is used to denoise in the detail weight of images. So the over-kill information phenomenon is avoided. The experimental results show that the algorithm has good effect of de-noising.(3) Analyze the quality maldistribution of the palm vein image. Because the enforcement effect of the equalization algorithm of the traditional local histogram is not obvious, an improved algorithm of it is proposed. This algorithm is applied to the Contourlet transformation of non-down-sampling and an enhancement algorithm of palm vein image is proposed. At last, the lines of the palm vein are enhanced by Gabor filter. The experimental results show that the enhancement effect of this algorithm is more obvious.更多还原