Abstract:

Medical image fusion technology plays an increasingly important role in clinical application because it contains multimodal image information. The fusion effect of medical image is more in line with human visual perception, reducing the influence of prior knowledge on the fusion effect and enhancing the detail expression. In this paper, a medical image fusion method based on Laplacian pyramid and convolutional neural network is proposed. Aiming at the problem of image artifacts, the regional Laplacian pyramid is adopted to improve the convolutional neural network in order to save more details and make parameters adaptive. The source images are input into the Laplacian pyramid for decomposition, and the improved convolutional neural network is used to generate the optimal weight graph to guide the fusion process. Finally, the fusion images are generated through the inverse process. Experimental results show that the method achieves good fusion effect on subjective vision and objective evaluation indexes.

Key words: medical image fusion, Laplacian pyramid, Convolutional Neural Network（CNN）

摘要：

医学图像融合技术因其包含多模态的图像信息，在临床应用中起着越来越重要的作用。医学图像融合效果符合人类视觉感知，减少先验知识对融合效果的影响和增强细节表现力一直是努力的方向。提出基于拉普拉斯金字塔和卷积神经网络的医学图像融合方法，针对图像伪影的问题采用区域拉普拉斯金字塔，为保存更多的细节信息并使参数自适应，对卷积神经网络进行改进。将源图像分别输入区域拉普拉斯金字塔进行分解，采用改进的卷积神经网络生成最优权重图指导融合过程，通过逆过程生成融合图像。实验结果表明，提出的方法在主观视觉和客观评价指标上都取得了良好的融合效果。

关键词: 医学图像融合, 拉普拉斯金字塔, 卷积神经网络