Ho Jun Kim

Ph.D. student at Graduate School of Advanced Imaging Science, Multimedia and Film, Chung-Ang University (CAU)

I’m a Ph.D. student in Image/3D Engineering at Chung-Ang University, where I am advised by Prof. Hak Gu Kim at the IRIS Lab. My research focuses on 3D reconstruction through Novel View Synthesis, with the goal of comprehensively leveraging visual data, language models, and 3D data within a unified framework.

I am broadly interested in advancing 3D scene understanding and reconstruction by integrating multimodal large language models into the Novel View Synthesis pipeline. In particular, my recent research focuses on Gaussian Splatting, pushing it toward higher-fidelity reconstruction and bringing it into Virtual Reality (VR) systems, while grounding it in the rich priors offered by visual and language models. Ultimately, my research aims to build 3D reconstruction systems that flexibly engage with diverse modalities to deliver immersive and faithful representations of the real world.

Publications

  • Language-Guided Scene Text Segmentation via Hierarchical Text Context Reasoning and Implicit Glyph Refinement
    Ho Jun Kim, Hak Gu Kim
    IEEE TIP (Under Review)

  • Leveraging Text Signed Distance Function Map for Boundary-Aware Guidance in Scene Text Segmentation [paper]
    Ho Jun Kim, Hak Gu Kim
    IEEE Access 2025

  • Unveiling the Potential of Multimodal Large Language Models for Scene Text Segmentation via Semantic-Enhanced Features [paper]
    Ho Jun Kim*, Hyung Kyu Kim*, Sangmin Lee, Hak Gu Kim (*: equal contribution)
    IEEE ICIPW 2024

Education

  • Ph.D. Candidate in Graduate School of Advanced Imaging Science, Multimedia & Film (GSAIM), Chung-Ang University (CAU), Sep. 2024 - Present
    Advisor: Prof. Hak Gu Kim

  • M.S. in Graduate School of Advanced Imaging Science, Multimedia & Film (GSAIM), Chung-Ang University (CAU), Sep. 2022 - Aug. 2024
    Advisor: Prof. Hak Gu Kim

  • B.S. in Industrial Engineering, Dankook University, Mar. 2016 - Feb. 2022

Projects

  • Development of 360 Omnidirectional Video-based High-Fidelity Egocentric 3D Representation Model and Quality Improvement for Reliable Intelligent Metaverse
    National Research Foundation of Korea

  • Acquisition of 3D Precise Information of Microstructure and Development of Authoring Technology for Ultra-High Precision Cultural Restoration
    Korea Creative Content Agency

  • Development of Scene Text Segmentation Technology for Better Text Understanding
    Electronics and Telecommunications Research Institute (ETRI)

  • Development of Scene Text Detection Robust against Shape Changes
    Electronics and Telecommunications Research Institute (ETRI)