Video Deblocking Using Multipath Deep Neural Networks

In this study, a video deblocking approach using multipath deep neural networks is proposed. The proposed approach contains temporal fusion subnet, variable-filter-size (VFS) subnet, and enhancement subnet. Video deblocking is performed via early fusion so that temporal correlations between adjacent video frames are employed. Based on the experimental results obtained in this study, in terms of two objective performance metrics and subjective evaluation, the performance of the proposed approach is better than those of four comparison approaches.