基于PCA和自适应区域方差的图像融合方法*

在对源图像进行提升小波变换的基础上，针对分解得到的低频分量和高频分量各自的特点，选取不同的融合规则，采用基于PCA和自适应区域方差的图像融合方法，即低频近似系数采用基于主元分析(PCA)加权法，高频细节系数采用自适应局部区域方差的融合方法，最后进行提升小波逆变换得到融合图像。实验结果表明，与传统算法相比，该算法不仅提高了信息量和清晰度，而且提高了融合图像与源图像的相关系数，降低了扭曲程度，有效地保留了源图像的细节信息，得到了清晰的融合图像，具有良好的目视效果。

Image fusion algorithm based on PCA & self-adaptive region variance

This paper proposed a novel image fusion algorithm based on PCA and self-adaptive local region variance according to the different characteristic of the coefficients of low frequency and high frequency after lifting wavelet transform on the original image, that was, used a weighted method depending on principal component analysis in the low frequency image, and selected an algorithm called self-adaptive local region variance as the guide to the high frequency images. At last, obtained the fused image by applying inverse lifting wavelet transform. Compared with other traditional methods based on wavelet transform, the experimental results show that this algorithm not only increases entropy and average gradient effectively, but also improves the correlation coefficient, reduces the degree of distortion, holds detail information of original images and provides good visual effects.