摘要
本文论述了实现图象自动拼接的新方法,针对拼接问题的特殊性,提出基于匹配的图象拼接应由以下四步完成:1.生成梯度图象以消除噪声的影响并提高配准精度。2.基于自学习阈值的快速匹配方法适用于大量图象的匹配。3.利用基团生长的方法从多幅图象的空间病态超定关系中确定拼接关系。4.边界处理中以距离为权的灰度均值调整消除了拼接处的阶梯。优点是:能自动完成多幅图象的空间定位,拼接。速度快,精度高,拼接处光滑过渡,避免了人为误差。
his paper presents a new method for image mosaic. Analysis of characteristics of mosaic showed that the registration should be completed as follows. (1) To overcome disturbance and to remove mismatch, a preprocessing for clearing and normalizing subpictures should be required. (2) Fast registration algorithm based on selflearning threshold is suitable for registration of a large number of images. (3) To determine the position of each subpicture in a mosaic picture by clique growth. (4) To eliminate gradient in mosaic image by adjustment of gray average weighted with distance.The merit of this method is that mosaic of multiple images is completed by computer automatically and human bias factors are avoided. The operation is fast and precise. It can also assemble pictures smoothly which is not available by manual method.
出处
《中国生物医学工程学报》
EI
CAS
CSCD
北大核心
1997年第4期316-322,共7页
Chinese Journal of Biomedical Engineering
基金
国家自然科学基金
关键词
图象拼接
匹配
边界处理
微循环
血管网络图
Image mosaic
Registration
Selflearning threshold
Clique growth
Adjustment of gray average