Research team of Professor Yoon Kuk-Jin developed a technique for confidence measure selection and cost modulation for robust stereo matching
Research team of Professor Yoon Kuk-Jin developed a technique for robust stereo matching which enables computing accurate disparity map in uncontrolled environment. Unlike indoor environments, it is hard to compute disparity map which is calculated as the difference between the pixel values of the two images since the image is either too bright or dark in uncontrolled outdoor environments. In this work, research team analyzes the characteristics of various confidence measures in the random forest framework, trains a random forest using the selected confidence measures, presents a confidence-based matching cost modulation scheme and applies the proposed modulation scheme to popularly used algorithms. The proposed methods exhibited accurate and robust results in public datasets and in challenging outdoor datasets. This paper was published in IEEE's Transactions on Pattern Analysis and Machine Intelligence (TPAMI) Volume: 41 Issue: 6, 2019. This work was supported by The Cross-Ministry Giga KOREA Project, Next-Generation Information Computing Development Program and Samsung Research Funding Center of Samsung Electronics.
Min-Gyu Park, Kuk-Jin Yoon, Learning and Selecting Confidence Measures for Robust Stereo Matching, IEEE Transactions on Pattern Analysis and Machine Intelligence, June 2019, Volume 41, Issue 6, pp 1397-1411
IEEE TPAMI is the best journal of PATTERN RECOGNITION and ARTIFICIAL INTELLIGENCE, which has the best IF out of IEEE journals in 2019.