基于深度卷积网络的多聚焦图像自动融合方法

Automatic Fusion Method of Multi-Focused Images Based on Deep Convolutional Networks

不同聚焦状态的图像融合易产生模糊的问题,因此提出基于深度卷积网络的多聚焦图像自动融合方法。首先检测图像的多聚焦区域,利用深度卷积网络选择图像块的大小,以便于后续的图像融合。选取好图像块大小后,通过小波变换进一步确定频子带系数。最后,结合深度卷积网络的自动融合方法,融合不同聚焦状态的图像,得到一幅完整、视觉效果好的融合图像。实验证明,在融合过程中使用深度卷积网络的多聚焦图像自动融合方法可以得到清晰的图像,具有良好的应用效果。

Image fusion with different focusing states is easy to produce blur.In this case,the automatic fusion method of multi-focusing images based on deep convolutional network is proposed.The multi-focus region of the image is first detected,and the deep convolutional network is used to select the size of the image block to facilitate subsequent image fusion.After selecting the size of the image block,the frequency subband coefficient can be further determined by the wavelet transform.Finally,combined with the automatic fusion method of deep convolutional network,the images of different focusing states are fused to get a complete fusion image with better visual effect.Experiments show that the multi-focus image automatic fusion method using deep convolutional network can get clear images in the fusion process and have good application effect.