A GAN-Based Video Intra Coding

Intra prediction is a vital part of the image/video coding framework, which designed to remove spatial redundancy within picture. Based on set predefined linear combinations, traditional intra cannot cope with blocks irregular textures. To tackle this drawback, in article, we propose Generative Adversarial Network (GAN)-based approach enhance accuracy. Specifically, superior non-linear fitting ability, well-trained generator GAN acts as mapping from adjacent reconstructed signals unit, implemented into both encoder and decoder. Simulation results show that for All-Intra configuration, our proposed algorithm achieves, average, 1.6% BD-rate cutback luminance components compared video reference software HM-16.15 outperforms previous similar works.