Perception Guided Image Manipulation

Published in APGV (study on depth ordering), TAP and TOG (subtle gaze direction), and the Visual Computer (motion and depth ordering of content in images), This dissertation presents a novel approach to image editing and manipulation where the goal is to explicitly trigger certain visual cues. We refer to this as perceptually meaningful image manipulation. Existing image editing approaches typically do not take human visual perception into account, whereas our approach relies heavily on the principles of human visual perception. We present computer-based techniques for simulating artistic control of apparent depth in an image, conveying a sense of motion in an image, and subtly directing a viewer\s gaze about an image. Our techniques combine elements of traditional art with research from various fields of science. Our image-based technique for simulating artistic control of apparent depth works by automatically adjusting color or luminance in specific regions of an image. The color and luminance adjustment is based on several simple rules that artists use to convey and manipulate apparent depth in their work. Our technique for conveying a sense of motion in a visually static image introduces spatial imprecision in the image plane. This is also patterned after work done by traditional artists. It has been suggested that spatially imprecise stimuli are processed differently by our foveal vision and our peripheral vision and that this difference in visual processing causes the image to appear to have a dynamic component. Finally, our subtle gaze directing technique exploits differences in visual acuity and processing speed between our foveal and peripheral vision. We present brief, subtle image-space modulations to the low acuity peripheral regions of the ?eld of view. These stimuli, which are detected quickly, attract the slower, high acuity foveal vision to fixate on the modulated regions

Download paper here

authors: Reynold Bailey

Authors: Reynold Bailey
Download Paper