Research on FRUC Algorithm Based on Improved U-Net

Abstract

Frame rate up up-conversion(FRUC), as a video post-processing technology, is of great help in improving video quality. The widely used frame rate improvement technology is based on the motion-compensated frame interpolation (MCFI). Although this method significantly improves video jitter and blur, there will still be problems such as block effects and holes. The paper proposes an improved U-Net frame rate improvement method that combines the U-Net and the Residual Neural Network (ResNet). The ResNet structure can effectively solve the problems of information loss, gradient disappearance and explosion during transmission. Combining these two networks and predicting the interpolation frames of the video sequences, such interpolation frames are closer to the original frames, and the predicted interpolation frames are better and effectively avoid problems such as block effects and holes. Experiments show that the algorithm in this paper is superior to other FRUC algorithms in the PSNR value of the interpolated frame.