摘要
图像配准技术是各种图像处理中必不可少的重要环节.文章提出一种面向对象的多尺度图像配准算法,该算法将面向对象技术和区域增长分割法相结合,得到理想的特征点;然后利用最小绝对值误差法,并采用双向匹配策略对特征点逐一配对,形成GCP集;最后利用放射变换和双线性插值法完成图像的纠正得到最终的配准图像.利用ENVISAT图像和Landsat图像获取的苏州某区域的相关数据进行实验并做了详细的客观评价,证明该算法切实可行,获取到的配准图像清晰,精确度达97%以上.
Image registration technique is an essential part in the image processing, this paper presents an object-oriented multi-scale image registration algorithm, the algorithm obtains the desired feature points on combined of object-oriented technology and regional growth meth- ods; then forms GCP sets, used of the absolute minimum error method and two-way matching feature points strategy on a one to feature point pairing; finally to complete image correct on used of radiation transformation and bilinear interpolation, we got the final registration image. The completion of the image to correct the image to get the final registration Finally, radiation trans- form and bilinear interpolation. Use Landsat image acquisition ENVISAT images and associated data in a region of Suzhou to experiment and do a detailed objective evaluation show that the algo- rithm is acquired registration image clarity, accuracy of more than 97%. ENVISAT images and Landsat images related data obtained in a region of Suzhou were experimented, and detailed ob- jective evaluation, proving the algorithm is feasible, obtained registration image clarity, accuracy of more than 97%.
出处
《太原师范学院学报（自然科学版）》
2015年第4期63-67,共5页
Journal of Taiyuan Normal University:Natural Science Edition
关键词
面向对象
多尺度
图像配准
object-oriented
multiscale
image registration
作者简介
贺养慧（1983-），女，山西朔州人，硕士，朔州师范高等专科学校数计系讲师，主要从事计算机应用技术研究．
