Can Region of Interest Coding Improve Overall Perceived Image Quality?

In this paper we investigate the possibility of improving the overall perception of image quality by preferentially coding certain regions of interest (ROI) in an image. Experiments are conducted utilising an automated algorithm for visual attention (VA) to detect the primary ROI(s) in an image, and then encoding the image using the maxshift algorithm of JPEG 2000. The results from a 2 alternative forced choice (2AFC) visual trial show that, while there is no overall preference for the ROI encoded images, there is an improvement in perceived image quality at low bit rates (below 0.25 bits per pixel). It is concluded that a perceived increase in overall image quality only occurs when the increase in quality of the ROI more than compensates for the corresponding decrease in quality of the image background (i.e., non ROI).