Date: 25 July 2008 (Friday)

Time: 10.30 - 11.30 am

Venue: Meeting Room C, Room Number S1-B1c-111.

School of Electrical and Electronic Engineering, Nanyang Technological University

(Light refreshments will be provided after the talk)

Title: Document Image Compression and Compressed Domain Document Processing.

Speaker: Dr. Utpal Garain, Indian Statistical Institute, Kolkata, India.

Abstract:

With the advent of digital libraries, research in document image compression and compressed domain document processing finds a new dimension. Document image compression serves an obvious practical purpose:
it conserves storage in a database of document images and thereby improves computer systems' performance by reducing accesses to storage devices, enhances utilization of available bandwidth in a network. Researchers have studied document image compression as a special case of general data compression for the last quarter century and several techniques and standards have been proposed. This talk, therefore, starts with revisiting the popular and frequently used algorithms for compression of scan-digitized document images. Standards like CCITT-IV, JBIG, and JBIG2 will be discussed giving some reasonable details. The role of symbolic compression in JBIG2 will be highlighted. Idea of residue coding will then be introduced and lossy and lossless versions of a JBIG2 compliant method are to be talked about. With increasing amount of document images being stored in compressed format, efficient information retrieval in the compressed domain has also become a major challenge for proficient designing of digital libraries and document viewing facility in resource-constraint systems like palm-held devices. The discussion on compressed domain pattern matching and retrieval is broadly based on JBIG2 compression method. Requirements for designing a complete compressed domain document image retrieval system would be outlined. The existing research in the related area are to surveyed and then the prevailing studies on accomplishing several IR (information retrieval) tasks for document images would be reviewed. Several applications like stopword spotting, word stemming, keyword spotting, duplicate detection, summarization, etc are to be discussed and research needed to realize these applications in the compressed domain would be presented.

Brief biography of the Speaker:

Utpal Garain received his bachelor and master degrees in computer science and engineering in 1994 and 1997, respectively from Jadavpur University, India and Ph.D. degree from Indian Statistical Institute in 2005. Dr. Garain started his career in software industry and later on, he joined Indian Statistical Institute where he is, at present, serving as an Assistant Professor. His research interest includes document image analysis including document compression and OCRs, technology development in Indian languages, machine learning, etc. Dr. Garain is one of the key scientists who first developed an Indic script (Devanagari-Hindi and Bengali) OCR system that was finally transferred to industry for commercialization. For his significant contribution in pattern recognition and its applications for DIA and language engineering, Dr. Garain received prestigious Young Engineer Award in 2006 from the Indian National Academy of Engineering (INAE).