数字图象变形技术研究

发布时间：2018-06-05 22:10

本文选题：数字 + 图象　；参考：《中国科学院研究生院（计算技术研究所）》1997年硕士论文

【摘要】：图象变形是研究将一幅数字图象平滑地变为另一幅数字图象的技术，它属于一种关键帧动画制作技术。近年来，在影视、广告和教育行业，图象变形技术取得了广泛的应用。图象变形技术基本上由三部分组成：图象特征的描述、翘曲的生成以及转变函数的设计。其中最主要的是图象翘曲的生成。图象翘曲由两方面构成，即空间变换和图象重取样。空间变换计算待翘曲图象中各象素点的位移。由于计算翘曲时所依据的控制元素(点或线段等，它们描述了图象特征)是不规则分布的，图象变形中涉及的空间变换一般可归类为散列数据插值。图象重取样是指将一幅取样图象从一个坐标系转换到另一个坐标系的处理。两坐标系之间通过空间变换的映射函数相联系。重取样的中心问题是插值核的选择。在比较和讨论了几种典型的变形方法之后，本文提出了一种新的图象变形方法—基于多边形分割的方法。其基本思想是根据图象特征，利用星形多边形对源和目标图象进行对应的分割。和其他几种变形方法相比，该方法能较好地描述图象特征，，变形效果较好，在速度和控制等方面有所改进。
[Abstract]:Image deformation is a technique to study the smooth transformation of one digital image into another digital image. It belongs to a key frame animation technology. In recent years, in film and television, advertising and education industry, image deformation technology has been widely used. Image deformation technology consists of three parts: the description of image features, the generation of warpage and the design of transformation function. One of the most important is the generation of image warpage. Image warping consists of two aspects: spatial transformation and image resampling. The spatial transformation is used to calculate the displacement of each pixel point in the image to be warped. Because the control elements (points or line segments) on which warpage is calculated are irregularly distributed, the spatial transformations involved in image deformation can be generally classified as hash data interpolation. Image resampling is the process of converting a sampled image from one coordinate system to another. The mapping functions of two coordinate systems are connected by space transformation. The central problem of resampling is the selection of interpolation kernels. After comparing and discussing several typical deformation methods, this paper presents a new method of image deformation, which is based on polygon segmentation. The basic idea is to segment the source and target image by star polygon according to the image features. Compared with other deformation methods, this method can describe the image features well, and the deformation effect is better, and the speed and control are improved.
【学位授予单位】：中国科学院研究生院（计算技术研究所）
【学位级别】：硕士
【学位授予年份】：1997
【分类号】：TP391.41

【共引文献】