Rate Distortion Perception

Rate-distortion-perception (RDP) theory explores the fundamental trade-offs between data compression (rate), reconstruction fidelity (distortion), and perceptual quality. Current research focuses on developing algorithms and models, including alternating minimization schemes and diffusion-based methods, to efficiently compute and achieve optimal RDP trade-offs for various data types, such as images and video, under different perception metrics (e.g., Kullback-Leibler divergence, Wasserstein distance). This work is significant for advancing both information theory and practical applications like image and video compression, enabling more efficient and perceptually pleasing data representation at various bitrates. The incorporation of conditional information and the exploration of the role of common randomness are also active areas of investigation.