Prof. Weisi Lin, Nanyang Technological University, Singapore (FIEEE, FIET, CEng, Hon. FSIET; Associate Chair (Research); School of Computer Science and Engineering)
Bio: Lin Weisi is
an active researcher in intelligent image processing, perception-based
signal modelling and assessment, video compression, and multimedia
communication. He had been the Lab Head, Visual Processing, in Institute for
Infocomm Research (I2R),Singapore. He is a Professor in School of Computer
Science and Engineering, Nanyang Technological University, where he also
served as the Associate Chair (Research).
He is a Fellow of IEEE and IET, and has been awarded Highly Cited Researcher
2019, 2020, 2021 and 2022 by Clarivate Analytics. He has elected as a
Distinguished Lecturer in both IEEE Circuits and Systems Society (2016-17)
and Asia-Pacific Signal and Information Processing Association (2012-13),
and given keynote/invited/tutorial/panel talks in 50+ international
conferences. He has been an Associate Editor for IEEE Trans. Neural Networks
Learn. Syst., IEEE Trans. Image Process., IEEE Trans. Circuits Syst. Video
Technol., IEEE Trans. Multim., IEEE Sig. Process. Lett., Quality and User
Experience, and J. Visual Commun. Image Represent. He also chaired the IEEE
MMTC QoE Interest Group (2012-2014); he has been a TP Chair for IEEE ICME
2013, QoMEX 2014, PV 2015, PCM 2012 and IEEE VCIP 2017, and is a General
Co-Chair for IEEE ICME 2025. He believes that good theory is practical, and
has delivered 10+ major systems and modules for industrial deployment with
the related technology developed.
Speech Title: Modeling Visual Signal Sensitivity for Human Perception, Machine Understanding, and Both
Abstract: Visual signal sensitivity, termed as Just-Noticeable-Difference (JND), refers to the minimum change of an image to be distinguished by the user. This talk firstly discusses how to model various visual signals (images, videos, graphics, and so on) for human users. Since humans have developed unique characteristics in perception over the long evolution, effective JND modeling facilitates optimization of user-centric system performance and utility of available resources (bandwidth, memory, battery, computation, device cost/size, etc). Machines are becoming the ultimate users for a rapidly increasing amount of visual signals in this AI era, and therefore as the second part of the talk, we will investigate into the concept of JND to be extended with machine tasks, including scenarios where both humans and machines are ultimate users. Finally, possible further research directions will be highlighted, inclusive of exploration toward true multimedia that consists of hearing, smell, touch and even taste aspects as well.