Remembering Thomas Huang: Image Processing and Computer Vision Pioneer

Every day 3.2 billion images and 720,000 hours of video content are shared on the internet. The quality of digital media today was largely influenced by computer scientist, electrical engineer, and writer Dr. Thomas Huang, who innovated the field of image processing.

 

Professor Thomas Huang. Image courtesy of the University of Illinois 

 

Optical signal processing, image compression, pattern recognition, and computer vision are among Huang's most important research and teaching areas. Huang is credited for pioneering the technologies behind Google Street View, Zoom backgrounds, and UHD TV.

But according to his own words, his most remarkable accomplishment was his mentorship. 

 

Thomas Huang's Career Mobility

Thomas (Shi-Tao) Huang, Ph.D., was born in Shanghai, China, in 1936. He earned his B.S. at the National Taiwan University (NTU), where he also met his wife Margaret. 

A world-renowned educator and researcher, Huang thought that an every-now-and-then job change is productive for one’s career. He never invested more than ten years in a job. 

Having taught and researched at MIT and Purdue, he spent his last years as a professor at the electrical and computer engineering (ECE) department at the University of Illinois. Specifically, he worked at the Image Formation and Processing (IFP) Lab and the interdisciplinary Beckman Institute for Advanced Science and Technology, which he founded alongside his fellow researchers.  

 

A Shift to Digital Image Processing

Initially, Huang was interested in network theory. However, during his graduate studies, he became interested in digital image processing, a technology that first appeared in the 1960s. His two theses on the subject of pictorial statistics and pictorial noise earned him his M.S. degree and his Ph.D. degree on the emerging topic.  

Huang published more than 1,300 research papers, authored more than 20 books, and mentored more than 100 students, many of whom followed in his footsteps as researchers or computer scientists. (A full list of research areas he worked on is available on his IEEE author’s profile page.) 

What made Huang one of a kind was his interdisciplinary approach to three research areas—signal processing, pattern recognition, and computer vision—all of which brought him outstanding accomplishments. 

 

Huang's Mark on Image Processing and Compression Standards

Huang was most interested in image compression. He worked on both facsimile (bi-level) and continuous tone images, setting the scene for international G3/G4 fax standards. 

He proposed block transform coding, which is a part of the JPEG standard for lossy image compression. Additionally, he worked on MPEG4 (advanced video coding) and studied pulse-coded modulation, wavelet image compression, and fractal image compression. 

He changed the way we store and retrieve images. Before his contributions, photo negatives and video cassettes were the only way to store images; he made it possible to develop compression standards for CDs and international video. 

He researched digital filters, the relationships of speech and image processing, multi-dimensional filter testing, efficient median filters, and least-squares fitting of two 3D point sets. 

He was the first to research multi-frame super-resolution, a method for enhancing the quality of low-resolution images for computer vision. He also created content-based image retrieval (CBRI), or the ability of search algorithms to retrieve similar visual content based on a specific image query.

 

Line-connected points of an original image based on the Haar-based wavelet salient point extraction algorithm. Image courtesy of the Image Formation and Processing Lab, University of Illinois

 

Laying the Groundwork for Computer Vision

Thomas Huang spoke enthusiastically about his work in computer vision, particularly in helping computers recognize human faces. The polymathic professor made important discoveries in digital holography, 3D modeling, and 3D motion analysis of the human face and body. 

One of his most enduring accomplishments was his discovery of relationship patterns between 3D movement and 2D images—a finding that is the basis for today’s Google Street View. 

Huang also developed a text messaging system with a talkative head that could deliver voice messages with content-appropriate gestures and tone. 

After his retirement, Huang shifted toward sparse representations, artificial intelligence, and deep learning algorithms for unlabeled image identification. He had great hope for computers to one day identify and categorize images without labels.    

 

Huang's Lasting Legacy

Thomas Huang received the most distinguished academic honors from the universities where he taught and international acclaim from science institutes. He became the first William Everitt Distinguished Professor at the University of Illinois and held the prestigious Maybelle Leland Swanlund Endowed Chair in Electrical and Computer Engineering. 

The long list of his professional associations includes election in the U.S. National Academy of Engineering (2001), foreign membership in the Chinese Academy of Sciences and the Chinese Academy of Engineering, and membership in Academia Sinica, a leading academic institution in Taipei City, Taiwan. 

In 2001, he received well-deserved recognition from IEEE with the IEEE Jack Kilby Signal Processing Medal. He is an IEEE Life Fellow, a recipient of the IS&T and SPIE Imaging Scientist of the Year Award, the King-Sun Fu Prize of the IAPR, and the Azriel Rosenfeld Life Time Achievement Award at ICCV.   

 

Tom and Margaret Huang in 1959. Image courtesy of Thomas Margaret Huang

 

Thomas Huang died peacefully in 2020 at the age of 83. He has left the legacy of two fellowships. The most recent one, the Thomas and Margaret Huang Ph.D. Fellowship in ECE, was established posthumously in June 2022, two years after his death, to support students of electrical engineering and computer science. A previously established fellowship, also attributed to him and his wife Margaret, was in the field of signal processing and data science.