Multi-focus Image Fusion Algorithm Based on Unsupervised Deep Learning

Abstract

With the development of deep learning, multi-focus image fusion based on deep learning has become a research hotspot. However, since this type of network does not have a large amount of labeled data for supervised training and generates training data that meet the actual situation of supervised learning, it is difficult to take advantage of supervised learning. To overcome this problem, an unsupervised learning method based on a convolutional neural network is proposed for multi-focus image fusion to generate all-focus images. This method uses a new CNN structure that can be trained for fusion without the need for true full-focus images and uses structural similarity and pixel loss to compute the loss. Experiments show that this method can preserve the effective information of the original image and make the fused image rich in detail and visual effect.