CityU student wins Best Paper Award for new image search method
Joan Yip
A Computer Science PhD student at City University of Hong Kong (CityU) has proposed a new data processing method that allows a user to search through billions of pictures, photos and images on the web at a faster rate.
Currently, one of the ways a user can search an image is to upload an image to a search engine and look for pages that include visually similar images. The image search is normally completed (literally) in the blink of an eye.
Tao Liang’s aim was to improve this high level of search efficiency. He combined two different approaches, namely, the “image hashing” and “Cauchy Graph” methods to propose a new data processing technique that enhances large-scale data analysis and thus results in a higher speed of image retrieval. “The search time is measured to the microsecond,” said Tao Liang.
This technique could prove useful to software or search engine companies. It also has the potential to be extended to multimedia searching of text, audio, video and other multimedia formats, added Tao Liang.
The image search method will help people solve problems in a different way. Tao Liang cited an example of how he used the image search method to identify a potted plant for which he did not know the name: “I simply took a picture of the plant and searched for web pages that included pictures similar to that of the plant.”
In December 2012, Tao Liang presented a paper on this technological breakthrough at the 2012 Pacific-Rim Conference on Multimedia. His paper was selected out of 80 papers for the Best Paper Award.
“After the presentation I received several requests for a copy of the presentation slides. Both my supervisor and I are happy that the audience was interested in our work,” said Tao Liang, thanking his supervisor, Professor Horace Ip Ho-shing, Acting Vice- President (Research and Technology) and Acting Dean of Graduate Studies, for all his encouragement and guidance.