Image deblurring processing model combining spatial domain and frequency domain information

The existing image deblurring methods usually use the spatial domain or frequency domain information directly to restore the clear image, ignoring the complementarity of spatial domain information and frequency domain information. The object structure can be recovered effectively by using the spatial domain information of the image, while the texture details can be recovered effectively through the frequency domain information. In this paper, a simple and effective image deblurring framework is proposed, which can make full use of the spatial domain and frequency domain information of the image to produce high quality clear images. First, two independent networks with the same structure are used to learn the mapping relationship between fuzzy image and clear image from spatial domain and frequency domain respectively. Then, a single fusion network is used to fully integrate the image information of the spatial domain and the frequency domain to further improve the quality of the clear image. The three networks are linked to form an end-to-end, learnable large network, and different networks influence each other and finally obtain high-quality clear images through joint optimization. Experimental results indicate the efficacy of the proposed image deblurring method, which fuses spatial domain and frequency domain information.