Compressed Video Quality Enhancement Method Based on Block Coding Features

To solve the issue that existing compressed video quality enhancement algorithms do not fully utilize the characteristics of compressed videos, the intrinsic relationship between video encoding and the task of compressed video quality enhancement was studied and a targeted non-aligned compressed video quality enhancement algorithm was designed contrapuntally, utilizing a three-dimensional convolutional neural network (3D-CNN). Experimental results show that compared with the high efficiency video coding (HEVC) standard, the peak signal-to-noise ratio (PSNR) of the proposed method is improved to 0.465 2 dB when low delay (LD) configuration and quantization parameter (QP) is 37. Compared with MGANet proposed in data compression conference (DCC), the PSNR increase of the proposed algorithm is improved by 15.1%.