Image Classification and Segmentation for Coding
[ Abstract ] [ Index ]
Abstract
More and more multi-media applications like distance learning and
teleconferencing require compression of images that contain both text and graphics.
In these applications text quality suffers dispraportionately when text and
graphics are subjected to the same compression scheme. In this
paper we show that there are large difference between text and graphics
and these can be exploited to achieve better coding performance. With this as
motivation we present methods found in the literature to segment the text
regions and graphics regions in images. Then we present new
methods that perform roughly 25% better than those we found in the
literature. Finally, we suggest enhancements and a sample application.
Index