Modelling Locally Disorderly Images using Fixation-Suppression

Abstract

This paper details an alternative approach to rendering 3D digital environments using fixation-based disorder fields that maintain high object saliency and provide greater depth perception across the entire visual field. The research undertaken compares the visual effect of applying disorder to an animated sequence as a simple 2D image space post-effect process, with a new 3D world space approach. Much research has been undertaken to establish how we perceive relative space when viewing 2D images. Extracting accurate 3D depth cues from 2D images can be difficult to assimilate, especially when camera position/settings are unknown. In optics, a sense of depth is achieved through the focal length. Converging light appears sharp and ‘in focus’, whilst poorly converging light appears blurry and ‘out of focus’. However, adding blur to an image has its drawbacks. More picture information is destroyed when the pixels of an image are blurred (averaged together), compared to when they are ‘scrambled’. It has been argued that whilst both images lose information, the scrambled image contains more information than the blurred one. Initial results in 2D screenspace identify key areas where local disorder decreases depth perception and creates visual confusion when applied to animated sequences. This paper proposes an alternative approach to control the disorder, which overcomes these problems when applied to the moving image. Keywords-disorder, blur, r-buffer, saccadic-suppression,