Abstract: Restoration method for lossy character image usually relies on low-pass filter, it turns the image smoothed, but meanwhile the characters’ edge will be fuzzier. Aimed at the defect of the traditional method, a new restoration approach based on the anisotropic diffusion partial differential equations is put forward in this paper, this approach can make the smooth course weaken at the perpendicular direction to the edge of lossy character, but strengthen at the parallel direction. Discrete algorithm is illustrated based on 8-neighborhoods of every element. Experiments indicate an encouraging high restoration rates compared to low-pass filter. In addition, calculation complexity of this algorithm has also been analysed.

Key words: Anisotropic, Image Restoration, Lossy Character, PDEs

摘要： 有损字符复原技术传统上以低通滤波为主，在平滑字符的同时，使得字符边缘更加模糊，甚至互相干涉．利用基于各向异性扩散的偏微分方程(PDEs)来对字符图像进行复原，使得字符图像的平滑过程在垂直于边缘的方向减弱，在平行于边缘的方向增强．基于每个像素点取8邻域，设计了这种新方法的离散算法，并对离散参数作了讨论．实验表明这种新的有损字符复原方法和传统的低通滤波方法相比，具有更好的复原效果，从而为后期的字符重建打下更好的基础．另外还针对这种算法的计算复杂度进行了讨论．

关键词: 各向异性, 图像复原, 有损字符, 偏微分方程