Gaze-J2K: gaze-influenced image coding using eye trackers and JPEG 2000

Abstract

The use of visual content in applications of the digital computer has increased dramatically with the advent of the Internet and world wide web. Image coding standards such as JPEG 2000 have been developed to provide scalable and progressive compression of imagery. Advances in image and video analysis are also making human-computer interaction multi-modal rather than through the use of a keyboard or mouse. An eye tracker is an example input device that can be used by an application that displays visual content to adapt to the viewer. Many features are required of the format to facilitate this adaptation, and some are already part of image coding standards such as JPEG 2000. This paper presents a system incorporating the use of eye tracking and JPEG 2000, called Gaze-J2K, to allow a customised encoding of an image by using a user’s gaze pattern. The gaze pattern is used to automatically determine and assign importance to fixated regions in an image, and subsequently constrain the encoding of the image to these regions.