Research

IML연구실에서는 영상 미디어 기반의 화질 개선 및 정보 센싱 기술을 연구하고 있습니다. 딥러닝 기법을 기존의 컴퓨터 비전 및 영상처리 기법에 응용하여 영상 미디어품질을 혁신하는 연구를 진행하고 있습니다.

고품질 영상 미디어 서비스 실현을 위한 다양한 연구활동을 하고 있으며, 인간 및 환경 친화적인 ICT 연구 개발로 우리 사회 및 인류에 공헌 하고자 합니다. 

연구자에게 가장 중요한 덕목은 새로운 것을 탐구하고 발견하는 능력과 그것을 알기 쉽게 표현하는 능력 입니다.

적극적으로 자신의 아이디어를 제시하고 도전하는 학생이 훌륭한 연구자로 성장할 수 있는 환경을 제공하고 있습니다. 

이에 관심이 많은 학생의 참여를 언제든지 환영합니다. 

Deep learning architecture for image fusion

We are studying an efficient deep learning algorithm to fuse heterogeneous images. In particular, we are developing a deep fusion architecture to compute the cross-correlation for the efficient fusion between RGB and NIR images.

Knowledge distillation architecture

Knowledge distillation has been populary studied for network complexity reduction. We are studying an effective knowledge distillation architecture as a prior for image enhancement. Also, its usefulness is studied for image segmentation with small dataset.

Multi-spectral imaging for future cameras

We are studying a novel approach for high-quality image acquisition by integrating multi-spectral RGB and NIR bands. 

It aims at enhancing visual quality  (color constancy, low light enhancement, and denoising) and improving vision performances (NIR-to-RGB conversion) in the camera ISP, particularly for dark illumination environments and bad weather conditions.

Hyperspectral imaging

Hyperpsectral imaging has been popularly used for various remote sensing fields such as agriculture. We are researching on spectral imaging specially for plant growth estimation. In particular, a deep learning algorithm is studied for hyperspectral recovery from RGB images, which is quite challenging ill-posed problem.

Image decomposition under AC lights

Using the sinusoidal variation of AC lights, the camera image is decomposed into illumination and reflectance components, based on the Dichromatic and the Retinex models, and they are applied to image enhancement techniques for digital cameras such as color constancy, highlight removal, and multi-exposure fusion.

AI for fast and phytochemical-enhanced plant production

We have launched a new pioneering research project recently, which focues on ICT and plant science. It aims at understanding bio-feedback mechanisms between plant and its environments using AI and vision technologies.

Our ultimate goal is to develop a vision based AI system to grow plants fast.

These technologies are applied to both indoor plant factory and outdoor smart farm.