二值和灰度图像的直接多尺度表示算法  

Direct algorithm of multiscale representation of binary and greyscale images.

作　　者：李将云[1] 

机构地区：[1]浙江大学计算机图像图形研究所,浙江杭州310027

出　　处：《浙江大学学报（理学版）》2006年第6期658-661,670,共5页Journal of Zhejiang University（Science Edition）

基　　金：国家自然科学基金资助项目(60473130);973发展规划基金资助(2004CB318000)

摘　　要：图像多尺度表示是图像处理的重要内容.常规的离散处理方法常依赖于坐标或网格的选取,因此处理结果很不稳定.本文提出了不依赖网格划分方法、不依赖坐标选取方式的快速、直接多尺度表示算法.应用该算法可直接得到任意指定一层(指定尺度)的图像,因而没有大多数算法因为逐层计算所带来的误差传播或误差扩散现象,同时该算法因保持处理前后图像的“整体”相似,因而能最大限度地体现形状保持乃至拓扑保持.算法计算非常简单,且不局限于特定类型的图像,可用于高要求的图像处理领域.Multiscale representation of images plays an important role in image processing, As general approaches to discrete images depend on coordinate systems or grid lines that are created within the source images, the destination images are unstable. A quick and direct algorithm for multiscale representation of images was presented which does not rely on the systems or lines mentioned above. An image with any scale can be obtained directly just from the source image if the algorithm is applied, so there is no error diffusing which can be usually found in many other scale-by-scale discrete image precessing algorithms. In addition, the method can give very similarity between the source and destination image, thus keeps their shape or topology. Also, there is a little computation in the simple algorithm which may be applied to a great diversity of source images in highly demanded image processing.

关 键 词：图像多尺度表示 图像放大 图像缩小 可视化 图像处理 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]