摘要
图像的小波多分辨表征是把图像特征按尺度和方向映射到由小波变换系数构成的金字塔结构的各层中.在此数据结构中,使用不同的基于区域的特征选择方法,实现了对各原始图像的明显特征的选择,包括基于能量判据和基于边缘检测的方法.并且结合这两种方法,综合利用了小波系数的方向信息来进行特征选择.实验结果表明,这些不同的信息融合途径,都能有效地实现基于像素级的多重图像融合,特别是有效地克服由于原始图像的灰度特性和边缘特性不相容对图像融合带来的困难.
The MRA(multiresolution analysis) of an image based on wavelet transform supplies a multichannel decomposition of an image. The image details are projected to different levels of the wavelet coefficients pyramid according to their scales. In this data structure, multiimage fusion is realized by several different area based feature selection methods, including the energy based criterion, edge detection method and using wavelet coefficient direction information in feature selection. It is shown that these kinds of algorithms are particularly effective to overcome the image fusion difficulties caused by the gray scale characteristics discrepancy and the feature inconsistency of the source images. 〖KH*2D]
出处
《北京理工大学学报》
EI
CAS
CSCD
1997年第4期458-463,共6页
Transactions of Beijing Institute of Technology
基金
国防科技预研基金
关键词
小波变换
多重图像融合
图像融合
图像处理
wavelet transform
multiimage fusion
area based feature selection