HIGH-FIDELITY GENERATIVE IMAGE COMPRESSION USING GAN’S AND XGBOOST APPROACH

---

Gadde Suresh 1, K.S.Thivya 2 & M.Anand 3

Vol 18, No 3 ( 2023 )   |  Licensing: CC 4.0   |   Pg no: 204-221   |   Published on: 30-03-2023

---

Abstract
Our research delved into the integration of Generative Adversarial Networks, XG Boost, and compression techniques to develop a sophisticated generative lossy compression system. The investigation covered various factors such as normalization layers, generator and discriminator architectures, training strategies, and perceptual losses. Our system produced visually appealing reconstructions similar to the original input, functioning effectively across a wide range of bitrates and even for high-resolution images. We evaluated the system's performance using several perceptual metrics and a user study, demonstrating our approach to be superior to existing methods, even at bitrates exceeding 2 x bitrate. In summary, our research successfully bridged the gap between rate-distortion-perception theory and practical implementation.

Keywords:
#

---

Download Full Article PDF