基于Contourlet域ICA和SVM的图像融合

Image fusion method based on contourlet-domain ICA and SVM

提出了一种基于Contourlet域独立分量分析(ICA)和支持向量机(SVM)的图像融合方法。首先对源图像进行Contourlet变换,再提取其高频系数的独立分量特征,并通过粒子群优化的SVM实现分类,最后进行图像重构得到融合结果。给出了实验结果,采用均方差(MSE)、信噪比(SNR)、信息熵(H)、空间频率(SF)、清晰度(CL)和相关系数(CR)等评价指标对融合效果进行了定量评价,并与加权平均法、基于Contourlet变换或基于ICA的图像融合方法进行了比较。结果表明,所提出的方法能取得更优越的融合效果。

An image fusion method based on eontourlet transform,independent component analysis（ICA） and support vector machine（SVM） is proposed. Firstly, the contourlet transform is used to perform a multi-scale decomposition of each image. Then, the ICA is used to extract the independent component features of the high-frequency components. And the SVM optimized by a particle swarm algorithm is trained to classify the fused image. Finally the fused coefficients are reconstructed to obtain the fusion results. The experimental results are evaluated quantitatively according to the evaluation items such as mean square error, signal to noise ratio, entropy of information, space frequency, definition, correlation coefficient and so on. The results are also compared with those of other image fusion methods based on weighting of average,eontourlet transform or ICA transform. It is shown that the image fusion method based on eontourlet transform and ICA can obtain superior results to others.