Learned Video Compression

Learned video compression uses deep learning to improve the efficiency of video encoding and decoding, aiming to surpass the performance of traditional codecs like H.266/VVC. Current research focuses on enhancing speed and efficiency through techniques such as low-resolution representation learning, improving motion compensation with methods like deformable convolutions and transformers, and optimizing prediction and reference quality adaptation. These advancements hold significant potential for reducing bandwidth requirements and improving video streaming quality across various applications.