Thomas Huang is one of the few remaining Beckman originals. He played a significant role in the preliminary building committees, and, when the doors opened in 1989, Huang eagerly moved his lab to Beckman, forming the Image Formation and Processing Group.
“There was definitely a need for the Beckman Institute. Problems were becoming more and more complicated in society, so it seemed necessary to have an interdisciplinary approach,” Huang said. “It really makes a difference when you get people physically together versus just remotely collaborating— work gets done.”
Huang’s career represents the interdisciplinary nature of the Institute. As an educator, mentor, and researcher, he has been awarded the highest recognition in three distinct fields: signal processing, pattern recognition, and computer vision—an enormous accomplishment, demonstrating his commitment to interdisciplinary research.
“This is an amazing aspect of his research,” said Zhi-Pei Liang, a co-chair of Integrative Imaging who has worked with Huang for many years. “Most researchers are happy to get to the top of a narrow field, but Tom has impacted three areas. That shows how influential his career has been.”
In addition to his research accomplishments, Huang has also mentored more than 100 Ph.D. students in his career.
“The one thing I have found over the years is—whatever I have accomplished, it’s entirely due to my students, so my career has really been the blessing of great students,” Huang said. “Mentoring students is the accomplishment I am most proud of. That’s one of the reasons I chose teaching as a career. I do research, I teach, but even if I don’t do well in research, I’m always educating students.”
His care for his students has not gone unnoticed. This past year, two Beckman student researchers were awarded the inaugural Thomas and Margaret Huang Award for Graduate Research, which was Aestablished by Huang’s former students James J. Kuch and Chang Wen Chen. The award honors the contributions Huang and his wife, Margaret, have made to science, engineering, and society at large.
“He is an inspiration to his students, not only through his lifetime achievement and his contributions to science, but also how he brings so much passion, energy, and creativity to the work,” Liang said. “He is a role model to many of us, and his work has impacted our society significantly, far beyond his papers, his books, and his many prestigious awards indicate.”
Huang’s career has spanned three major universities. He received his Doctor of Science (Sc.D.) in 1963 from MIT, and then stayed at MIT as a faculty member for 10 years. In 1973 he moved to Purdue for seven years, and then moved to the University of Illinois in 1980.
“My original intention was to stay in one place for about 10 years and then move to another place to avoid becoming stale, but then I was trapped here by the Beckman Institute,” Huang said, with a laugh. “The environment here was too good—I couldn’t possibly move anywhere else.”
Throughout his career, Huang has made a wide range of pioneering and fundamental contributions to image and signal processing, pattern recognition, computer vision, multimedia, and human-computer interface. He has contributed to 21 books, more than 600 journal and conference papers, and has been elected to the National Academy of Engineering (USA), Chinese Academy of Engineering, Chinese Academy of Sciences, and Academia Sinica (Taiwan).
The one thing I have found over the years is—whatever I have accomplished, it’s entirely due to my students, so my career has really been the blessing of great students. Mentoring students is the accomplishment I am most proud of. That’s one of the reasons I chose teaching as a career. I do research, I teach, but even if I don’t do well in research, I’m always educating students. - Tom Huang
Huang’s visionary work early on in his career helped shape current practices in imaging. Before Huang’s work, there were very few ways to store an image: photographic negatives and video cassettes. His work was instrumental in developing compression standards for CDs, for example.
“Because of Tom’s pioneering work, there are now a seemingly endless numbers of ways to capture, store, and share images,” Liang said. “He has contributed more than anyone else to the technical underpinning of current international fax, image, and video compression standards. Without these standards, it simply would not be possible for us to store and transmit the huge amounts of multimedia data that all of us encounter on a daily basis.”
In pattern recognition and computer vision, Huang’s contribution when he first came to the U of I was his creative formulation and solution of the problems of 3D motion estimation from 2D image sequences, a long-standing problem in the field at the time. 3D motion estimation has had many important applications, including navigation and orientation in the 3D world, video coding, and object tracking. Recent advances in 3D urban modeling programs, such as Google’s StreetView, have foundations in Huang’s work in the 1980s and 1990s.
His paper titled “Image Retrieval: Current Techniques, Promising Directions, and Open Issues” in the Journal of Visual Communication and Image Representation, received the “Most Cited Paper of the Decade Award” in 2010. It has been cited 564 times by other articles in the journal and more than 1,000 times in Google Scholar since first appearing in 1999.
Huang continues to build collaborations and develop research projects across a wide spectrum at Beckman.
“We are concentrating on three areas right now: big data, deep neural net learning, and high-performance computing, and the three seem to come together,” Huang said.
One of the projects he and his students are working on is focusing on the human-computer interface, especially in emotional recognition in education and learning.
They’re working to build algorithms that can read the emotions of people through their facial expressions, as well as tell their age, ethnicity, and gender. This tool could be effective in online learning environments.
“We’re developing a framework to create a responsive, real-time online education experience,” said Huang. “If you have an online computer learning system where the computer is interacting with the students, what the computer does should be based on the emotion and cognitive state of the student. We’re trying to estimate that by using noninvasive methods—so by the facial expressions picked up by a webcam on the computer, for example.”
If the student uses a computer as a tutor, this real-time feedback would allow the computer, at each interchange, to decide what to say next, with this decision based on the state of the student—not just what the student is saying, but how the student looks: puzzled, bored, etc.
Despite all the success he’s had in his five decades of research, Huang remains a humble leader within the Beckman Institute, the University of Illinois, and society at large.
“His technical accomplishments are extraordinary, his impact on the field is amazing, but he is an even better person and colleague,” Liang said. “He is an enormous strength to this institution.”
“The biggest change in the last 25 years (mostly in the last five years) in my research is the emergence of the triad: big data, deep learning, and high-performance computing. Ours is the age of big data, which is characterized by the 3V model: volume, variety, and velocity.
“The approach of deep learning, especially deep neural networks (DNN), appears very effective for analyzing big data. On the other hand, the training of DNN requires huge amounts of data, which could be provided by big data available on the web.
“All this is impossible without high-performance computing, since high data volume and rapid rate of change make big data analysis extremely computationally intensive. Thus, the triad of big data, deep learning, and high-performance computing are truly made for each other.
“I foresee that the future of my field will continue to be shaped by these three areas. Additionally, in the next 25 years, my prediction and hope is that there will be large-scale collaborative research linking closely neuroscience and machine learning, leading to a deep understanding of how the human brain works in processing and understanding multimodal data.”