Work in this theme includes the development of new algorithms for image and video coding and compression as well as the study of automated video annotation and synthesis. Compression and coding has used ideas from fractal geometry for representing the self similarity within images. Novel fractal features have been explored for image classification. New methods for raster to vector conversion have also been explored with special application to cartoon images and video. In the area of video annotation and synthesis joint audio-visual processing has been used for extracting meaningful information from video sources. Robust and efficient point feature extraction algorithms have been developed for video tracking, image registration and mosaicing applications. Recent work has also been aimed towards using visual interpretation as part of an automated robot guidance system.
Other work in this area includes people-tracking through the use of camera position-based geometric models of human objects, for applications in security and surveillance monitoring.