文档介绍：河北工业大学硕士学位论文

基于灰度和几何特征的图像匹配算法研究

摘要

图像匹配是计算机视觉和图像处理领域一项重要的研究工作,本文主要对图像匹配领
域进行了深入细致的研究。
本文主要研究了两类图像匹配算法:基于灰度信息的算法和基于特征的匹配算法。对
基于灰度信息的算法,主要研究了两种算法,一种是对传统算法进行研究,提出了一种
改进的互相关匹配算法,另一种是根据图像编码的思想,对图像进行分块,重点研究了基
于灰度值编码的匹配方法。实验表明,此算法在遥感图像和工件字符定位方面,尤其是在
复杂背景(目标与背景难分离)下,算法都具有很强的鲁棒性和稳定性。对基于特征的匹配
算法,主要研究了 HU 不变矩、圆形度、矩形度等几何特征,最后选取几种特征作为特征
参数,运用基于欧式距离的匹配方法进行匹配. 对二百多个样本图像进行测试,匹配成功
率达到了 %。得到了很好的实验效果。

关键词:计算机视觉模板匹配图像处理互相关

i
基于灰度和几何特征的图像匹配算法研究
The Study of Image Matching Algorithms Based On Gray Value and
Geometric Features

ABSTRACT

Image matching is an important research topic puter vision and image processing. A
great deal of work is done in the field of image matching in the paper.
It studied two types of image matching algorithms in the paper. gray-scale
information-based algorithm and feature-based matching algorithm. About the algorithm
based on the information of gray–scale, it mainly studied two algorithms, Firstly it studied
the traditional algorithm, Then one new improved cross-correlation algorithm was proposed. The
other is based on the thinking of image coding. This algorithm divided the image into certain
size blocks called R-block. It focused on the algorithms of gray value image coding. Through t
he experiments, It is found that the algorithm had a very strong robustness and stability ,When it
was used in remote sensing images and Optical Character positioning, particularly in the
Complex background(with the background of the difficult goal of separation). About the
feature-based matching algorithm, Firstly, it mainly studied HU invariant moments,
elongated-ness, roundness and other geometric features. Finally it selected a few features from
them as the features of the parameters. Then it was calculated and matched with the algorithm
based on template matching of Euclidean distance. Two hundred of samples was tested and
experimented, and the average accuracy rate is %.