Notice
We are recruiting student researchers for the Photorealistic Immersive Scene Media Laboratory (PRISM Lab.). [Lab Introduction Slides].
If you are interested, please contact jongbeomjeong@ksnu.ac.kr.
2026.02: Our lab member Seungmi Lee won the President’s Award at the 1st Hansung University Physical AI Hackathon! [News]
2026.01: Our standardization contribution was adopted at the 153rd MPEG standardization meeting. Our technology will be applied to TMCV, the candidate reference software for the MPEG Gaussian Splat Coding (GSC) standard. [News]
2026.01: Yuseong Choi from our lab was selected for the KSNU AIX-Boost program. He will receive KRW 1,000,000 in support during the project period.
2026.01: Our lab was awarded the KSNU R-Seed program (a sole award in the AI/SW field), and we will receive equipment support worth KRW 20,000,000. [News]
2025.10: Our lab was selected for the NIPA GPU support program. We can use NVIDIA H100 GPU resources (worth approximately KRW 40,000,000). [News]
Photorealistic Immersive Scene Media Laboratory (PRISM Lab.), Kunsan National University (KSNU)
PRISM Lab. aims to solve real-world problems across media capture, compression, transmission, and rendering for representing real spaces with high immersion by incorporating state-of-the-art AI technologies. Beyond conventional 2D media, photorealistic immersive media expanded to 3D can provide experiences that transcend time and space—not only for humans but also for robots and machines. To realize media that is difficult to distinguish from reality, many researchers continue to tackle diverse challenges. Our lab focuses on international standardization that can be adopted globally, while also building practical systems that address real needs.
Large-Scale Real-Space Acquisition
We study capture technologies to replicate large-scale spaces—such as buildings or entire campuses—into high-precision digital representations. We systematically combine multiple sensing modalities (LiDAR, multi-camera systems, drones, and more) to collect wide-area spatial data without missing regions. A key focus is automated preprocessing, including effective noise removal and precise alignment across different coordinate systems. Through this, we build high-resolution digital twins of large real spaces and secure high-quality, large-scale datasets that serve as the foundation for immersive media services.
Hyperimmersive 3D Real-Space Representation
2차원 평면 영상의 한계를 극복하고 사용자가 공간 안에서 자유롭게 이동하며 관찰할 수 있는 6DoF (6자유도) 기반의 초실감 표현 기술을 다룹니다. 전통적인 포인트 클라우드 (point cloud) 기법부터 인공지능 기반의 신경망 방사형 필드 (NeRF) 와 3D 가우시안 스플래팅 (3DGS) 까지, 공간의 질감과 조명을 정밀하게 재현하는 최신 기법들을 깊이 있게 연구합니다. 특히 3DGS 기술을 고도화 및 인공지능 기반 기술을 결합하여 실시간 렌더링 성능과 photorealism을 동시에 확보하는 최적의 모델링 기법을 연구합니. 이러한 연구를 통해 사용자가 마치 실제 현장에 있는 듯한 깊은 몰입감을 느낄 수 있는 차세대 미디어의 시각적 원천 기술을 개발합니다.
MPEG Gaussian Splat Coding (GSC) Standardization
To enable interoperable and broadly usable next-generation 3D media, we actively participate in MPEG Gaussian Splat Coding (GSC) standardization. Since 3D Gaussian data is massive, innovative compression and efficient transmission technologies are essential. We research key technologies for coding and transport, focusing on practical streaming scenarios such as low-latency delivery and bitrate control algorithms driven by real-world requirements. By contributing to international standards, we aim to strengthen global competitiveness and help build a robust ecosystem for immersive media platforms.
Location
Kunsan National University (KSNU): (54150) Digital Information Bldg., Rm. 336, 558, Daehak-ro, Gunsan-si, Jeonbuk-do, South Korea