International Display Workshops General Incorporated Association

11:10 AM - 11:30 AM

[VHF3-2] Image Compression and Restoration Using Deep Learning Considering Spatial Frequency Characteristics of the Visual System

Naoki Tada1, Keita Hirai1, *Takahiko Horiuchi1 (1.Chiba University (Japan))

Image coding, JPEG, U-Net, CNN, Visual system

This paper proposes image compression and restoration techniques that consider the visual characteristics of humans with respect to the spatial frequency. The method reduces information with low visual sensitivity, encodes and decodes by JPEG, and restores the high-quality image using U-Net. The feasibility of the method was verified experimentally.