基于CNN的非下采样剪切波域多聚焦图像融合

摘要

结合非下采样剪切波变换的时频分离优良特性,提出了一种基于卷积神经网络( convolutional neural networks,CNN)的非下采样剪切波变换(non-subsampled shearlet transform,NSST)域图像融合算法.首先对源图像进行NSST分解,其次对分解的低频系数进行基于CNN的融合策略.最后对分解的高频系数进行基于向导滤波(guided filtering,GF)的改进加权的拉普拉斯能量和( improved weighted sum of Laplace energy, IWSML)模取大融合策略,然后将根据不同融合规则融合后的频率系数进行NSST反变化获取输出的清晰目标图像.实验结果表明,该方法不仅可以获得更利于人眼接受的视觉效果图,且有效地提高了融合图像的客观性能评价指标.%In this paper, a new multi-focus image fusion algorithm was proposed based on convolution neural network (CNN) in non-subsampled Shearlet ( NSST) domain by using the advantages of time-frequency of NSST. Firstly, the source image was decomposed by NSST. Secondly, the fusion strategy based on the convo-lution neural network was applied to the low frequency coefficients of the decomposition. Then, the improved weighted sum of Laplace energy ( IWSML) based on the guided filtering ( GF) were carried out to the high-frequency coefficients of the decomposition. Finally, the fused image could be gotten by inverse NSST trans-form. Experimental results showed that the fusion algorithm could not only achieve better visual effects, but al-so improve its objective evaluation index.