Research

Research Directions

We study intelligent visual information processing and transmission through four closely connected directions spanning visual perception, intelligent coding, 3D computing, and AI-powered visual applications.

Overview

Intelligent Visual Information Processing and Transmission

The iMediaCodec Lab focuses on intelligent visual information processing and transmission across perception, compression, computing, communication, and presentation. Our work is organized into four major directions: visual perception modeling, intelligent visual coding and communication, 3D visual coding and computing, and artificial intelligence with visual applications. Together, they connect visual quality understanding and efficient media representation with immersive content processing and intelligent visual systems.

Visual Perception Modeling Intelligent Visual Coding & Communication 3D Visual Coding & Computing AI-Powered Visual Applications

Direction 01

Visual Perception Modeling

Grounded in the physiological and psychophysical mechanisms of the human visual system, we develop unified models that connect visual perception mechanisms with computational prediction. Our research covers image quality assessment (IQA), video quality assessment (VQA), and point cloud quality assessment under compression distortion, transmission degradation, and immersive viewing conditions. These perceptual models further support coding parameter optimization, resource allocation, and quality-of-experience improvement.

Visual Mechanisms & Modeling Image Quality Assessment (IQA) Video Quality Assessment (VQA) Point Cloud Quality Assessment

Human visual perception mechanism — Human Visual Perception Mechanism

Subjective and objective point cloud quality assessment workflow — Subjective and Objective Point Cloud Quality Assessment

Direction 02

Intelligent Visual Coding and Communication

To serve the distinct requirements of human viewing and machine analysis, we study efficient compression and reliable transmission of images, videos, and multimodal visual data. Key topics include learned image compression (LIC), neural video coding (NVC), and image compression for machine vision, together with perception-driven rate-distortion optimization, semantic feature coding, adaptive rate control, and visual communication over complex networks.

Learned Image Compression (LIC) Neural Video Coding (NVC) Image Compression for Machine Vision Joint Perception-Coding-Communication

Video coding optimization framework — Video Coding Optimization Framework

Conventional and end-to-end video compression frameworks — Conventional and End-to-End Video Compression

Direction 03

3D Visual Coding and Computing

We study efficient representation, transmission, and interaction for 3D scenes and immersive media, covering the acquisition, coding, enhancement, reconstruction, and rendering of 3D video, point clouds, light fields, and multiview video. Our work includes joint geometry-attribute compression, spatiotemporal modeling of dynamic 3D content, virtual and augmented reality (VR/AR), and 3D Gaussian representation and coding for six-degrees-of-freedom (6DoF) experiences, live streaming, and remote interaction.

3D Video Point Cloud Coding & Enhancement VR / AR 3D Gaussian Techniques

3D video coding framework — 3D Video Coding Framework

VR video live streaming system — VR Video Live Streaming System

Direction 04

Artificial Intelligence and Visual Applications

We develop AI-driven methods for visual content analysis, understanding, generation, and interaction. Our research spans image enhancement and restoration under complex degradations, short- and long-term video understanding and generation, cross-modal vision-language modeling, and intelligent visual systems for real-world environments. By combining generative AI with visual foundation models, we advance applications in smart cities, industrial inspection, remote sensing, and immersive media.

Image Enhancement & Restoration Video Understanding & Generation Vision-Language Modeling Intelligent Visual Systems

Frontier applications of artificial intelligence — Frontier AI Applications

Intelligent video analysis — Intelligent Video Analysis

Open Positions

We continuously recruit Ph.D. students, master's students, and postdoctoral researchers. Outstanding undergraduates interested in our research are also welcome to join the lab.

Ph.D. Students Master's Students Postdoctoral Researchers Undergraduate Research

Contact

We welcome academic and industry collaboration in intelligent visual coding, visual quality assessment, 3D media, VR/AR, and AI-powered visual applications.

zhangyun2@mail.sysu.edu.cn