Currently, I am studying PhD on "New Approaches to Object-Based Image Segmentation Using Fuzzy Clustering" which is the most important and challenging tasks for MPEG-4 for real images under the supervision of Dr. Gour Chandra karmakar, Lecturer, GSCIT, Monash University and Prof. Laurence S. Dooley, Multimedia Chair, IT faculty, Monash University.

The segmentation results of any clustering algorithm are solely dependent on the type of the features used and the information about the objects in an image. This raises an open question about which type of feature produces better results for which type of object. Even though a set of features used in clustering algorithms does not produce better results for all images. This limits the generalization of clustering algorithm. To address these issues, a new algorithm and motivates to combine the results individually produced by any clustering algorithm considering each feature type.

However, it does not produce expected results for the images containing similar objects with respect to pixel intensity and surface variations and the objects are connected with each other because of the arbitrariness of FCM using pixel location. In this case the potentially of the segmentation techniques would be improved if the shape specific information of an object could be incorporated. Considering this issue,  an algorithm is being developed by incorporating general information of an object. Note that this algorithm may be able to solve the two most difficult problems in the multimedia research area – the segmentation of the video object plane (VOP) of real video for object-based video coding in MPEG-4, which has already been implemented for only a synthetic video and object-based content retrieval in MPEG-7. It is also expected to apply this algorithm for the implementation of both MPEG-4.

Designed by Ameer Ali.  Please post your comments to-

 Ameer.Ali@infotech.monash.edu.au, ameer7302002@yahoo.com.