Deep learning based video compression techniques with future research issues

The advancements in the domain of video coding technologies are tremendously fluctuating in recent years. As the public got acquainted with the creation and availability of videos through internet boom and video acquisition devices including mobile phones, camera etc., the necessity of video compression become crucial. The resolution variance (4 K, 2 K etc.), framerate, display is some of the features that glorifies the importance of compression. Improving compression ratio with better efficiency and quality was the focus and it has many stumbling blocks to achieve it. The era of artificial intelligence, neural network, and especially deep learning provided light in the path of video processing area, particularly in compression. The paper mainly focuses on a precise, organized, meticulous review of the impact of deep learning on video compression. The content adaptivity quality of deep learning marks its importance in video compression to traditional signal processing. The development of intelligent and self-trained steps in video compression with deep learning is reviewed in detail. The relevant and noteworthy work that arose in each step of compression is inculcated in this paper. A detailed survey in the development of intra- prediction, inter-prediction, in-loop filtering, quantization, and entropy coding in hand with deep learning techniques are pointed along with envisages ideas in each field. The future scope of enhancement in various stages of compression and relevant research scope to explore with Deep Learning is emphasized.