• 学院公告

学术报告通知
作者:null 发布日期:2011-10-10 浏览次数:

学术报告:A Revisit to the Fundamentals of Image Processing

时间:10月14日(星期五) 上午 10:00am

地点:郁B107

报告人:澳大利亚Ngai Kwok博士

Dr N. M. Kwok

School of Mechanical and Manufacturing Engineering

The University of New South Wales, NSW, Australia

 

Abstract

With advances in imaging devices and technologies, the application of image processing in a large number of engineering processes has found an ever increasing pace. In order to provide further developments and better understandings of image processing, this presentation is focused on issues and studies in the context of the fundamental processes. These include contrast enhancement, image segmentation and colour correction. Despite the promising results obtained from current solutions of these problems, there are still some approaches initiated from alternative perspectives that have not yet been fully explored. Furthermore, the feasibility of incorporating intelligent computation methodologies is attempted.

In the area of colour image contrast enhancement, studies are made to tackle the problem of artefacts generated from conventional histogram equalization where groups of pixel intensities are changed to discrete levels and the limitations found in a global equalization paradigm. A possible solution is suggested to allocate pixel intensities in a continuous spectrum obtained from a smoothed version of the original image histogram. Furthermore, equalization is made by dividing the image into sectors and results are derived from a fusion of the individually enhanced sectors. The popular un-sharp masking and bi-literal filter algorithms use pre-defined masks to obtain the enhanced image. An attempt is made with the use of the particle swarm optimization routine to tune the mask characteristics such that an image of improved contrast can be delivered.

A popular approach in colour image segmentation is to convert the image into its gray level equivalent and thresholds are applied with some criteria such as maximizing the inter-segment separation or the information content of the segmented image. For the former direction of approach, the effects of conducting segmentation in transformed colour spaces are evaluated. The incorporation of computational intelligence is investigated where the cellular automata approach is adopted in segmenting the image into super-pixels. Instead of needing to pre-specific the number of segments, this method iteratively obtains satisfactory segmentation results.

It is well known that illuminations have a profound effect on the representation of the original colour of an imaged object. Conventional colour correction methods, such as those based on the gray-world assumption, frequently destroy the information conveyed by the image intensity that describes the scene brightness. An approach based on gamma correction, using a bi-sectional search for the optimal gamma value, is developed such that the original scene brightness can be preserved or transformed to any user specified value.

Further research directions are identified. In particular, the use of computation intelligence for optimization purposes has been found promising.

 

Biography

Dr Kwok received the PhD from the University of Technology Sydney, Australia, in May 2007. His thesis topic was on the simultaneous localization and mapping for a mobile robot using a single camera. He was employed as a post-doctorial research fellow in the same university working on non-linear modelling, smart building control, perception acquisition and robotic manipulator control. Since Nov 2008, Dr Kwok became a lecturer at the School of Mechanical and Manufacturing Engineering, The University of New South Wales, Australia. His research interest includes image processing, intelligent computation and control engineering.  His current projects include the development of mobile robot coordination systems, crowd behaviour monitoring, segmenting remote sensing images, and characterization of imaged soot particle in a diesel combustion engine. Dr Kwok has about 100 publications, 45 listed in SCI with 113 citations and 66 listed in SCOPUS with 193 citations. He has been the regular reviewer of IEEE Transactions on Automation Science and Engineering, Engineering Applications of Artificial Intelligence, Computers and Electrical Engineering. He also served as Programme Committee members and session chairs in a number of international conferences such as the IEEE CASE 2010, IEEE CISP 2010 and 2011. He was the recipient of the best paper award at the 2007 IEEE International Conference in Automation Science and Engineering. Dr Kwok is a Member of the IEEE, Chartered Engineer, Member of the Institution of Engineering and Technology.