Video Interframe Image Generation Based On Spatial Continuity Generative Adversarial Networks

Traditional algorithms for increasing the frame rate have the advantage of eliminating video motion jitter and smearing,but at the same time may cause the image edges to be unclear.This paper proposes a method for generating images of video frames based on Spatial Continuity Generative Adversarial Networks to solve the low frame rate video playback is not smooth and the problem of edge blur brought by the use of traditional methods to improve video frame rate.The SC-GAN model is trained by using video-based image datasets and inter-frame images are generated using image spatial continuity.First of all,during the preprocessing of video data,the low frame rate video is divided into frames and sequentially numbered to form a dataset.In the training of the model,this data set will be used to train the SC-GAN model.Second,the SC-GAN model uses an auto-encoder as a discriminator to replace the convolutional neural network structure in the original GAN discriminator in this paper.The advantage of this method is that it can carry out the feature learning of the sample quickly and improve the convergence speed of the sample training.The SC-GAN model introduces two concepts of Wasserstein distance and balance parameters.In this paper,the loss function based on Wasserstein distance is used to optimize the network model.At the same time,a balance parameter γ is introduced to keep the balance between generator and discriminator in the training process,which effectively avoids the problem of model collapse.Finally,using the spatial continuity of video frame images,this paper uses Adam's optimization algorithm to find an optimal distribution between two adjacent frames and maps them to the image space to obtain the inter-frame images.In order to illustrate the SCGAN model's ability to generate,we also use PSNR and SSIM to evaluate the quality of the generated inter-frame images.From the results generated,it solves the problem of unclear edge caused by the traditional method and obtains better quality image frames,which provides a novel way for improving the frame rate of the video.