Image Processing and Computer Vision

• Content Based Image Retrieval refers to the algorithms and techniques for querying image databases based on the objects, colours and features found in the images rather than on text attributes such as the image filename, dimensions, etc. I am interested in exploring colour and colour spaces and am working currently on databases of remote sensing images.
• Forensic Applications of image processing are gaining currency, especially in the analysis of questioned documents (whose authenticity is in doubt). We are investigating techniques for detecting erasures, obliterations and alterations on questioned documents.

Optical Character Recognition

Along with my colleague, Atul Negi, DRISHTI, a complete OCR system for Telugu, was developed in the department. It is probably the first and the only complete OCR system for Telugu. I am greatly interested in analyzing confusion matrices, the development of new features based on the characteristics of Telugu script, and in binarization, character and word segmentation problems.