Versatile Image Enhancement

Recent development in visual SLAM includes a various application. However, we encounter images are degraded easily with noise from the haze, fog, and smoke.

1) Real-time channel invariant dehazing We provide an effective software module that is applicable for dehazing. Conventional approaches for dehazing usually exploit color information to estimated required parameters. Our solution is fully real-time (>10Hz) and works equally well for both color and grayscale images. Our method is applicable to various degradation, fog, smoke, and turbid underwater.

Click to see video

Younggun Cho and Ayoung Kim, Visibility Enhancement for Underwater Visual SLAM based on Underwater Light Scattering Model ICRA 2017

Click to see video

Younggun Cho, Ramavtar Malav, Gaurav Pandey and Ayoung Kim, DehazeGAN: Simultaneous Hazing and Dehazing Networks using Unpaired Image-to-image Translation (IROS 2017)

2) Exposure control for robust visual SLAM Vision based robotics often encounter either under or over exposed situation. Typical auto-exposure control scheme focuses on overall image quality for everyday life. Edges and detail enhancement is rather important for the robotics application.

Click to see video

Joowan Kim, Younggun Cho and Ayoung Kim, Exposure Control using Bayesian Optimization based on Entropy Weighted Image Gradient ICRA 2018

Related Projects

  • [NRF 2015 - 2018]
    Real-time sonar and optical image enhancement for in-water structure monitoring using underwater robot
    (수중 로봇의 인프라 모니터링을 위한 광학센서와 소나센서의 강건성과 실시간성 연구)
Copyright © 2014-2020 IRAP
Page last modified on May 31, 2018, at 05:59 PM