基于改进的SIFT算法的图像配准技术研究与应用

发布时间：2018-03-09 12:26

本文选题：图像配准　切入点：SIFT　出处：《青岛科技大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着计算机技术的飞速发展,图像配准已成为图像处理领域的研究热点,它在医学、遥感、军事、计算机视觉等领域得到广泛的应用。随着应用的深入,对图像配准的适应性、实时性和准确性都提出了更高的要求。近几年出现了很多图像配准算法,取得了许多研究成果,但是仍存在一些问题需要解决和完善,所以对图像配准算法进行深入的研究有着十分重要的意义。本文重点研究了基于局部特征的SIFT算法,目的是提高配准算法的实时性和准确性,主要研究内容包括以下几个方面:首先,本文对图像配准的基础理论和关键技术进行了系统的研究。从图像配准的研究状况出发,分析了图像配准的基本框架,研究了配准算法的四种变换模型以及常用的三种重采样技术,并从复杂度、鲁棒性、配准精度和配准时间四个方面对算法的性能进行评价与分析。其次,详细的研究了SIFT算法的基本原理和实现过程,针对算法在特征提取时存在复杂度高、实时性差的问题,提出了一种结合ORB算法的SIFT改进算法,降低原算法的时间复杂度,提高算法的实时性。在配准阶段采用了改进的Kd-树最近邻查询算法,减少原算法在“回溯”操作时消耗的大量时间,结合RANSAC算法对匹配对进行精简,提高算法的准确性。将改进后的算法与原算法在视角变化、尺度变化、旋转变化三个方面进行对比与分析。从实验数据来看,改进后算法在配准速度和配准精度方面都得到提升,证明了改进后算法的可行性。最后,本文将改进后的算法应用到场景图像的拼接中。采用对比度增强和直方图均衡化的方法对图像进行预处理,并将改进后算法应用到场景图像的配准,在图像融合阶段分析了三种消除拼接缝隙的方法,经对比发现加权平滑算法能实现重叠区域的平滑过渡,之后给出了实际的拼接效果图。
[Abstract]:With the rapid development of computer technology, image registration has become a hot research topic in the field of image processing, it is in the medical, military, remote sensing, widely used in the field of computer vision. With the deepening of the application of image registration, adaptability, real-time and accuracy are put forward higher requirements. In recent years there has been a lot of image registration algorithm, obtained a lot of achievements, but there are still some problems to be solved and improved, so in-depth study of image registration algorithm has a very important significance. This paper focuses on the research of SIFT algorithm based on local features, the purpose is to improve the real-time and accuracy of the registration algorithm, the main research contents include the following: first of all, the basic theory and the key technology of image registration are studied. Starting from the research status of image registration, image registration The basic framework of four transformation model registration algorithm and three common resampling technique, and the complexity of the four aspects of robustness and precision of registration time on the performance of the algorithm is evaluated and analyzed. Secondly, a detailed study of the basic principle and implementation process of SIFT algorithm, the algorithm there is high complexity in the feature extraction, bad real-time problems, this paper proposed an improved algorithm based on ORB algorithm of SIFT, reduce the time complexity of the original algorithm to improve the real-time performance of the algorithm. At the registration stage using the improved Kd- tree algorithm of nearest neighbor query, reduce the amount of time consumed by the original algorithm backtracking "operation, combined with the RANSAC algorithm for matching to streamline, improve the algorithm accuracy. The scale change of the improved algorithm and the original algorithm in view of change, for comparison and analysis. The three aspects from the change of rotation The experimental data, the improved algorithm in the registration speed and accuracy have been improved, proved that the improved algorithm is feasible. Finally, the improved algorithm is applied to image stitching. By contrast enhancement method and histogram equalization of image preprocessing, and improved algorithm applied to image registration, image fusion in the phase analysis of three kinds of methods to eliminate the splicing gap, we found that the weighted smoothing algorithm can achieve a smooth transition of the overlapping area, given the actual splicing effect diagram.

【学位授予单位】：青岛科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】