My research interests include Multimedia Databases and Indexing, Data Mining and Database Management Systems.

Project 1:
Mutidimensional Index Structures for Multimedia Data Supporting Content-Based Information Retrievals

With the recent advancement of hardware, storing a large amount of image, video and audio data has become quite common, which emphasizes the growing needs of efficient multimedia organization and retrieval systems. The huge size of multimedia data makes indexing a crucial component for fast and efficient retrieval processes. Content Based Image Retrieval (CBIR) with Relevance Feedback (RF) has been a popular method for image retrieval. Hence, the need of an efficient index structure for images supporting popular access mechanisms like CBIR arises. The main challenge of implementing such an index structure is to make it capable of handling high-level image relationships easily and efficiently during access. The existing multidimensional index structures support CBIR by translating the content-similarity measurement into feature-level equivalence, which is very difficult and can often result in erroneous interpretation of users' perception of similarity. We propose a novel indexing and access method, called Affinity Hybrid Tree (AH-Tree) to organize large image data sets efficiently and to support popular image access mechanisms like CBIR by embedding the high level semantic image-relationship in the access mechanism as it is. AH-Tree combines Space-Based and Distance-based indexing techniques to form a hybrid structure which is efficient in terms of computational overhead and fairly accurate in producing query results close to human perception. SImilarity queries like Range and k-NN search are implemented. The proposed tree is compared in terms of comptation and query relevance with M-tree, a distance based index structures. To the best of our knowledge, the AH-Tree is the first attempt to combine the feature based and distance based multidimensional indexing techniques to introduce the best characteristics of each. I further extended the AH-Tree structure to support the other most popular multimedia data type, Videos, into a hierarchical framework called Hirarchical Affinity Hybrid Tree (HAH-Tree). It supports content-based video retrieval strategies for various video units considering both the inter as well as intra-unit relationships. Such retrieval techniques are implemented while following the k-NN criteria. Later, I proposed a generalized index structure called GeM-Tree which can support both images and videos along with their varied representations and retrievals strategies from within one single framework.

Related Publications:

• Kasturi Chatterjee and Shu-Ching Chen, " Affinity Hybrid Tree: An Indexing Technique for Content-Based Image Retrieval in Multimedia Databases," IEEE International Symposium on Multimedia (ISM2006), pp. 47-54, December 11-13, 2006, San Diego, CA, USA.
  Best Paper Award, IEEE International Symposium on Multimedia(ISM2006), December 11-13, 2006, San Diego, CA, USA.
• Kasturi Chatterjee and Shu-Ching Chen, "A Novel Indexing and Access Mechanism using Affinity Hybrid Tree for Content-Based Image Retrieval in Multimedia Databases," International Journal of Semantic Computing (IJSC), Vol. 1, Issue 2, pp. 147-170, June 2007.
• Kasturi Chatterjee and Shu-Ching Chen, "Hierarchical Affinity-Hybrid Tree: A Multidimensional Index Structure to Organize Videos and Support Content-Based Retrievals," accepted for publication, Proceedings of the 2008 IEEE International Conference on Information Reuse and Integration (IEEE IRI-08), Hilton Hotel, Las Vegas, USA, July 13 - 15, 2008.
• Kasturi Chatterjee and Shu-Ching Chen, "GeM-Tree: Towards a Generalized Multidimensional Index Structure Supporting Image and Video Retrieval," accepted for publication, the Fourth IEEE International Workshop on Multimedia Information Processing and Retrieval (MIPR2008), in conjunction with IEEE International Symposium on Multimedia (ISM2008), Berkeley, California, USA, December 15-17, 2008.
• Kasturi Chatterjee and Shu-Ching Chen, “A Multimedia Data Management Approach with GeM-Tree,” accepted for publication with minor revision, Journal of Multimedia.
• Kasturi Chatterjee and Shu-Ching Chen, “Towards a Multidimensional Index Structure Supporting Different Video Modeling Approaches in a Video Database Management System,” under second round of review for the special issue of International Journal of Information and Decision Sciences (IJIDS).
• Kasturi Chatterjee S. Masoud Sadjadi, and Shu-Ching C, "A Distributed Multimedia Data Management over the Grid," Multimedia Services in Intelligent Environments - Integrated Systems, Springer, 2009, accepted for publication.

Project 2:
Embedding Query Refinement in Multi-dimensional Index Structures Supporting CBIR

The relevance of the query results for CBIR depends largely upon successful semantic interpretation of image relationship. It is very difficult to capture the users' perception of similarity resulting in executuion of several iterations of queries. Query Refinement/Reformulation is an attempt to move the query point as close to the user choice as possible to increase result relevance. The naive approach is to treat a refined query just like a starting query and execute it from the scratch. But, it has been observed that refined queries are not modified drastically from one iterations to another. Hence, the execution cost can be improved dramatically by appropriately exploiting information and result set generated in the previous iteration. The goal is to optimize I/O and CPU usage. The existing implementation of multipoint query execution incorporating the query refinement technique in the multi dimensional index structures use arbitrary ditance function which cannot be supported by any distance-based index structure. Though feature-based index structures is able to address the problem, but they have the limilation of embedding high-level (semantic) image relationship "as it is" and use an error-prone and difficult way to translate the users' perception to low level feature equivalence. Our current work aims at solving this problem by enabling Query Refinement strategies at the distance based structure, thus fulfilling the two main goals:
1) Implement Query Refinement/Reformulation in the index structure efficiently to improve Query Result and at the same time reduce computation overhead
2) Enable the proposed AH-Tree (in prj 1) to handle the Query Refinemnt strategies so as to allow the high-level image relationship play a crucial role in Query Reconstruction/Multi-point Query mechanisms

Related Publications:

• Kasturi Chatterjee and Shu-Ching Chen, “Hybrid Query Refinement: A Strategy for a Distance Based Index Structure to Refine Multimedia Queries,” under preparation.

Project 3:
Multimedia Data Management in a Collaborative Environment

This project utilizes the social network concepts to understand multimedia data relations in a Collaborative Environment. Such understanding is leveraged and utilized in taking design decisions of various components of a database like the index structure, query processors, etc. This research is still in its early investigations stage.

Related Publications:

• Kasturi Chatterjee and Shu-Ching Chen, “Multimedia Data Management in a Collaborative Environment Utilising Social Network Concepts,” under preparation.