中文摘要：

目的：针对传统的图像增强算法中存在弱边缘增强效果差、同时噪声抑制较弱等问题，本文提出的一种无抽样方向滤波器组，并将其用在医学图像增强中。方法：首先，将一维低通滤波器转换成二维低通滤波器，此滤波器再经平移、旋转等操作得到其他多方向无抽样频域滤波器，并将各频域滤波器转化成空间模板，便于使用；其次，将多尺度分析方法与无抽样方向滤波器组结合对图像进行分解得到各子带图像，对各子带图像进行统计特性分析，确定图像的强边缘、弱边缘和噪声：最后，对此三类信息不同的线性变换分别进行处理，获得增强后的图像。结果：由于无抽样方向滤波器能够较好的捕获图像的方向信息，多尺度分析能够较好区分不同的边缘信息，所以该方法的结果很好的保护了图像的边缘及细节，同时有效的去除了图像的噪声，层次结构清晰，视觉效果有了显著改善。结论：将无抽样方向滤波器与多尺度分析结合．并根据不同纹理信息的特点，对各类信息采用不同的处理，是一种行之有效的图像增强方法。

英文摘要：

Objective： Aiming at enhancing weak edges of images and minimizing image noises not existed in classical methods, we present a decimation-free directional filter banks, and use them to enhance medical images. Method： First, all decimation-free directional filters are obtained by shifting and rotating the 2-D low-pass filter, which is developed from the 1-D low-pass filter. Furthermore, we decompose the input image by combining multiscale analysis with directional analysis to get all subband images. Then we study the statistics of all subband images to classify the image information, such as strong edge, weak edge and noises. Finally, we apply the non-linear transform to the three-class information to get enhanced image. Result： As the decimation-free directional filters have the ability in capturing directional information and multiscale analysis can distinguish from different texture information, the method preserves the edges and textures in images, moves noises efficiently and makes the image have clear structures and excellent visual quality. Conclusion： It is an effective method of image enhancement that combines the decimation-free directional filter banks with multiscale analysis and uses the non-linear transform to process the image according to the characteristics of different texture information.