Perceptual depth estimation from a single 2D image based on visual perception theory

Abstract

The depth of image is conventionally defined as the distance between the corresponding scene point of the image and the pinhole of the camera, which is not harmony with the depth perception of human vision. In this paper we define a new perceptual depth of image which is perceived by human vision. The traditional computation models of image depth are all based on the physical imaging model, which ignore the human depth perception. This paper presents a novel computation model based on the visual perception theory. In this approach, we can get the relative perceptual depth from a single 2-D image. Experimental results show that our model is effective and corresponds to the human perception. © Springer-Verlag Berlin Heidelberg 2006.