WONG, Tien-Tsin (T. T.) 黃田津

Professor
SMACM, MIEEE, FHKIE
Department of Data Science & Artificial Intelligence,
Faculty of Information Technology,
Monash University

He graduated from the Chinese University of Hong Kong in 1992 with a B.Sc. degree in Computer Science. He obtained his M.Phil. and Ph.D. degrees in Computer Science from the same university in 1994 and 1998 respectively. From August 1998 to August 1999, he was a visiting assistant professor in the Computer Science Department of Hong Kong University of Science & Technology. From August 1999 to June 2024, he was a professor in the Computer Science & Engineering Department of the Chinese University of Hong Kong. Since July 2024, he joined the Department of Data Science & Artificial Intelligence, Monash University as a professor. He received the IEEE Transactions on Multimedia Prize Paper Award 2005 and the Young Researcher Award 2004.

He was a member in Academic Committee of Microsoft Digital Cartoon and Animation Laboratory in Beijing Film Academy, visiting professor in School of Computer Science and Technology at Tianjin University, and the visiting research professor in Biomedical Engineering Department of Shanghai Jiaotong University. In the summer of 2000, he was a visiting researcher in Microsoft Research, Asia. He is the member of ACM (Senior Member), ACM SIGGRAPH, IEEE, and HKIE (Fellow). He is the Associate Editor of IEEE Transactions on Visualization & Computer Graphics, and Computer Graphics Forum. He has actively involved (as Program Co-chair, Program Committee and Organizing Committee) in several international conferences, including SIGGRAPH (2019, 2020, 2024), SIGGRAPH Asia (2009, 2010, 2012, 2013, 2018), Eurographics (2007, 2008, 2009, 2011, 2019, 2024), Pacific Graphics (2000-2005, 2007-2019, 2021), i3D (2010-2013), IEEE Virtual Reality (2011), ICCV (2009), Computer Graphics International (2004, 2006, 2012-2024), CVM (2012, 2013, 2015-2023), CAD/Graphics and Chinagraph, etc. Besides, he was also active in transferring graphics technologies to games industry, including writing articles in books for game developers (Graphics Gems V, Graphics Programming Methods, Shader X³, ShaderX⁴, ShaderX⁵, ShaderX⁶, and ShaderX⁷).

His main research interests include generative AI, computer graphics, computer vision, computational manga, image-based rendering, GPU techniques, medical visualization, and multimedia compression. Click [favorite] to see his favorite works he published in SIGGRAPH, CVPR, ICCV, ECCV, Game Developers Conference, and other venues.

Summary of His Research Contributions:

Invertible Generative Model In SIGGRAPH Asia 2018, we published a general neural network framework for solving invertibility problems. It contains a pair of encoding and decoding subnetworks for generating the desired visualizable content that can invert back to its original input. The encoding network encodes information into a visualizable latent, such that the information can be decoded via a decoding network. The visualizable latent can be controled via the objective function or a GAN. Due to the flexibility and generality of our model, it can be flexibily applied to many applications. We coin it as Invertible Generative Model. This framework can be applied for information hiding, data compression, and solving backward compatibility issues. For example, [color ↔ grayscale] we can hide the color information of a color image in an invertible grayscale (SIGGRAPH Asia 2018), so that this grayscale image can restore its color. [color ↔ halftone] Our framework is further challenged by extending it to invertible halftone (ICCV 2021), as the halftone image has a significant smaller solution space and its blue-noise requirement. [stereo ↔ mono] We can turn a binocular video into an ordinary monocular video (mononizing binocular videos, SIGGRAPH Asia 2020) for full backward compatibility and higher compression efficiency. [video ↔ image] We can also produce a real live photo (video snapshot, IEEE TPAMI 2020)), a normal still picture that can really animate, instead of storing a video sequence for a photo.

Computational Manga and Anime Motivated by painstaking manual process of manga/anime production, Tien-Tsin Wong has started a new field of computational manga and anime since 2005. The goal is to develop technologies for intelligently usage of computers in manga/anime production. These technologies include manga colorization (SIGGRAPH 2006), cartoon deringing (ACM TOG 2006), structure-aware halftoning (SIGGRAPH 2008), manga screening (SIGGRAPH Asia 2008), anime stereoscopization (SIGGRAPH Asia 2013), closure-aware sketch simplification (SIGGRAPH Asia 2015), manga screen removal (SIGGRAPH 2017), two-stage sketch colorization (SIGGRAPH Asia 2018), manga filling with screenVAE (SIGGRAPH Asia 2020). and the generative cartoon interpolation, ToonCrafter (SIGGRAPH Asia 2024). Manga colorization facilitates the conversion of original black-and-white manga into color manga via a few scribbles. Hence it saves the color manga production cost, and the style of converted color manga preserves the same style as the black-and-white version. Structure-aware halftoning and manga screening provide ways to turn the color photos into a b/w manga background in a fully automatic manner. This frees up the effort and time of manga artists, so that they can focus on the design of foreground characters. The recent ToonCrafter pushes one step forward towards automatic keyframes inbetweening by exploiting the motion priors from video diffusion model. It has been well received by the community users. These technologies have already drawn the attention of companies in the industry and has been reported by Hong Kong and international media.

Image-based Relighting He pioneered in the field of precomputed lighting. In 1997, he published the earliest work in the field of image-based relighting and proposed the model of apparent BRDF (ABRDF) (EGWR 1997). It is further generalized to plenoptic illumination function (IEEE TMM, 2002). This work is cited by many researchers and further extended by institutions worldwide. It has been taught in the advanced graphics courses offered by several renown universities worldwide. In 2005, his paper on plenoptic illumination function is further recognized and awarded IEEE Transactions on Multimedia Prize Paper Award 2005. Besides the representation, he also proposed practical encoding and rendering solutions by encoding the data with spherical harmonics (IEEE TCSVT, 2003), spherical radial basis function (SRBF) (IEEE TIP, 2006 and GDC 2007), PCA (IEEE TCSVT, 2005), and spherical wavelet, so that game sceneries with ultra-realistic lighting effects can be rendered in real-time. Hence the techniques can be applied in computer games development.

Weathering Simulation He also pioneered in the field of weathering simulation. In 1995, he proposed to determine the surface exposure to simulate dust accumulation (IEEE CGA, 1995) and other weathering phenomena (EGWR 1997), such as rusting, patina, and peeling. This surface exposure is referred as the first ambient occlusion, which becomes a popular technique to improve the photorealism of synthetic images, in computer graphics. With his contribution, the synthetic imagery becomes more realistic, because our real-world is full of imperfection or weathering. His work on weathering simulation is also reported by the best-selling computer graphics magazine (Randall Warniers, "Dirty Pictures", in Computer Graphics World, June 1998, pp. 50-60). His recent work of gamma-ton tracing (SIGGRAPH 2005) provides an even more comprehensive framework for simulating complex weathering phenomena.

Sphere Maps He has proposed and published a series of work on uniform sampling and/or partitioning of sphere. Sphere map is an important tool for rendering and many applications in other areas. By equally sampling/partitioning the sphere, we can efficiently represent any spherical function and fit it nicely in the computer memory. The techniques he has published include, unicube (IEEE TVCG), isocube (IEEE TVCG, 2007), rhombic dodecahedron Map (IEEE TMM), spherical Q2-tree (EGSR 2005), uniform cubemap, Hammersley and Halton points on sphere (Journal of Graphics Tools, 1997). This stream of work is presented in the Game Developers Conference 2007 , an influential conference in the game industry.

GPGPU Another significant contribution of Tien-Tsin Wong is in the field of GPGPU. He is one of the earliest researchers on pursuing GPGPU (general purpose computing on GPU). He proposed the first genetic algorithm on GPU (IEEE Intelligent Systems, 2007) and the earliest GPU shader realization of discrete wavelet transform (IEEE TMM, 2007). He also proposed the first high-quality pseudo-random number generator (PRNG) on GPU (Shader X5) (early GPU has no bitwise operators equipped, making it difficult to implement traditional PRNG). His work in GPGPU has significantly sped up the classical algorithms that are usually referred as computational intensive and slow in performance. Hence this makes these algorithms more practical.

Chinese Visible Human Another his contribution is in the Chinese Visible Human project. In this joint project, five bodies are sliced into thousands of cross-sectional images. The one with the highest data resolution requires more than 1T of storage. Compression and real-time rendering of this massive data is challenging. With the advances in GPU techniques, they have developed a real-time volume rendering algorithm on ordinary PC. The technology hence overcomes the cost barrier which impedes the wide usage of such massive data. The developed technology has been patented. Besides, he has also contributed in medical visualization projects including virtual endoscopy, virtual arthroscopy surgical simulation, and virtual Chinese acupuncture.

(last updated: June 2024)

WONG, Tien-Tsin (T. T.)

Summary of His Research Contributions:

Research Interests

Academic Awards

Publications

Technical Slides

Professional Service