Cloud-Mobile Video Super-Resolution Method Based on Deep Learning

With the rapid popularity of short video applications and the rise of streaming media platforms, video images have become an important medium for people's daily communication, and users' requirements for image quality are getting higher and higher. In order to improve the image quality of mobile devices or web applications, the Image Super-Resolution (SR) algorithm is often used to help improve the user experience. In order to effectively solve the problems of image deterioration caused by limited cloud-mobile transmission bandwidth, difficult model deployment, high energy consumption, and slow inference speed, this paper proposes a lightweight image super-resolution model based on attention mechanism and multi-scale residual network. First of all, the multi-scale characterization ability of the algorithm model is enhanced by combining dense connection and channel attention mechanism to improve the performance of image super-score reconstruction. Secondly, the loss function of mean variance and total variation is used in the training process to eliminate the impact of noise in the reconstruction process. Finally, through experimental analysis and comparison of results, it is concluded that the proposed cloud-mobile video super-resolution method has better image reconstruction performance and faster processing speed.