摘要
提出了一种快速提取多尺度融合系数的融合规则用于图像融合。规则应用于源图像进行多尺度分解,对各尺度上融合系数的提取采用基于邻域窗口的融合方式,低频采用改进的邻域熵提取匹配测度,高频采用跨尺度的邻域梯度来提取匹配测度,并给出了融合系数公式。基于图像融合对平移不变和方向选择的敏感性,提出结合双树复小波变换对图像进行多尺度分解新算法。实验表明,采用新算法融合的图像具有相对融合信息熵、标准差大、融合效果好等特点。
In order to achieve a high degree of accuracy on image fusion and make the algorithm simplified, a rapid extraction of multi-scale fusion coefficients rule is proposed and applied to image fusion. A multi-scale decomposition is carried out in the presence of the original image, and then the extraction of fusion coefficient on each scale is obtained according to neighborhood window, which adopts an advanced neighborhood entropy method to extract the match degree evaluation for low frequency part, and a trans-dimension neighDorhood gradient method for high frequency part. Together with that, the fusion coefficient formula is presented. Considering the sensitivity of translation invariance and direction selectivity on image fusion, the proposed algorithm adopts dual-tree complex wavelet transform for multi-scale decomposition. The experimental application shows the fusion images possess comparatively fusion information entropy and greater standard deviation, and high quality.
出处
《武汉理工大学学报》
CAS
CSCD
北大核心
2009年第9期104-106,共3页
Journal of Wuhan University of Technology
关键词
图像融合
双树复小波分析
融合规则
imagine fusion
dual tree complex wavelet transform
fusion ruler
