Dr. Li is a full professor in geo-informatics (cartography/remote sensing/GIS). He obtained his BSc in photogrammetry and remote sensing from Southwestern Jiaotong University (China) in 1982 and his Ph.D. from the University of Glasgow (UK) in 1990. Since 1990, he had worked at the University of Newcastle upon Tyne as a postdoctoral research associate, the University of Southampton and Technical University of Berlin (Germany) as a postdoctoral research fellow. In 1994, took up a lectureship at Curtin University of Technology (Australia). He joined the Hong Kong Polytechnic University in 1996 as an assistant professor. He was promoted to associate professor in 1998 and to professor in 2003.
Image fusion is emerging as a vital technology in many applications. Various measures are available for assessing image fusion quality. A survey shows that a total of nearly 30 measures has been in use. It has been found that some of them are highly correlated and these measures can be classified into difference-based, noise-based, similarity-based and information-based. In information-cased category, Shannon Entropy is in use but it has been found not very effective because it only takes into the consideration of the composition of the image pixels but doesn’t take into consideration of the configuration information. This leads us to re-visit the Boltzmann Entropy, which is classical measure of disorder in thermodynamics proposed by physicist Ludwig Boltzmann in 1870s, as this entropy takes into consideration of both composition and configuration information of images.
In this presentation, first of all, existing measures for assessing quality of image fusion will briefly reviewed; then the problems associated with Shannon entropy will be demonstrated and the difficulty of solving the Boltzmann equation (i.e. to define a good macrostate and then to determine the number of microstates) will be illustrated; then a feasible solution developed by the presenter and his collaborators (i.e. to define the macrostate of an image as its abstract or upscaled and to define the number of microstates as all the possible ways of downscaling from the macrostate to the original) will be introduced. Experimental results show this solution works very well. This opens a new door for research in image science in general and fusion quality in specific.