The Use of Image Blur as a Depth Cue

Abstract

Images of three-dimensional scenes inevitably contain regions that are spatially blurred by differing amounts, owing to depth-of-focus limitations in the imaging apparatus. Recent perceptual data indicate that this blur variation acts as an effective cue to depth: if one image region contains sharply focused texture, and another contains blurred texture, then the two regions may be perceived at different depths, even in the absence of other depth cues. Calculations based on the optical properties of the human eye have shown that variation in blur as a function of depth follows the same course as variation in binocular disparity with depth. Computational modelling has shown that the effect of blur on single-step edges is very similar to its effect on random fractal patterns, because the two stimuli have similar Fourier amplitude spectra. Blur discrimination thresholds for the two stimuli were also very similar, and could be predicted by a model based on high-spatial-frequency discrimination. A comparison of blur discrimination thresholds with the range of binocular stereopsis indicates that blur and disparity cues cover different distance ranges: stereopsis is most effective for distances relatively close to fixation, while blur information should be more effective for larger distances.