Temporal Integration of Spatially Filtered Visual Images

Abstract

Factors which govern the temporal integration of spatial information were examined in a group of five experiments. A series of high-pass and low-pass spatially filtered versions of a visual scene were generated. Observers' ratings of these filtered versions of the scene for perceived image quality indicated that quality was determined both by the bandwidth of spatial information and the presence of high-spatial-frequency edge information. When sequences of three different versions of the scene were presented over an interval of 120 ms the perceived quality of the resulting composite image was determined both from the ratings of the individual components of that sequence and from the order in which these components were presented. When the order of spatial information in a sequence moved from coarse to fine detail the perceived quality of the composite image was significantly better than when the order moved from fine to coarse. This evidence of a coarse-to-fine bias in pattern integration was further investigated with a detection paradigm. The pattern of errors once again indicated that temporal integration of spatial information was superior when a coarse-to-fine mode of information delivery was employed. Taken together the data indicate that the pattern-integration mechanism has an inherent order bias and does not accumulate spatial information so efficiently when the 'natural' coarse-to-fine order is violated.

Face perception: An integrative review of the role of spatial frequencies

“…For the anisotropic-integration hypothesis, the integration order is fixed and the majority of empirical data indicates that LSFs are integrated faster than HSFs (Hughes, Fendrich & ReuterLorenz, 1990;Hughes, Nozawa & Kitterle, 1996;LaGasse, 1993;Parker, Lishman & Hughes, 1992Watt, 1987Watt, , 1988, though one study found that HSFs were integrated faster (McSorley & Findlay, 2002). Bachmann (1987) showed two target images (an eye with an eyebrow and a face) mixed up with eight images of different objects.…”

“…Although Parker, Lishman and Hughes (1992) found empirical support for the anisotropic hypothesis using a procedure that involved displaying a three-image sequence containing different frequencies of the same scene, shown for an exposure time of 40 ms, they also argued that the task (judging the quality of an image) was not precise enough and it probably would have been better to ask to the subjects to perform direct discriminations. Hence, Parker, Lishman and Hughes (1997) tested subjects' ability to classify full-bandwidth images into coarse-to-fine and fine-to-coarse sequences containing distorted versions of the originals, i.e.…”

Diagnostic Colors Mediate Scene Recognition

“…In higher-level vision, numerous studies have examined how luminance cues supported the recognition of faces (Breitmeyer, 1984;Costen, Parker, & Craw, 1994;Fiorentini, Maffei, & Sandini, 1983;Sergent, 1986), objects (Parker et al, 1996), and scenes Parker, Lishman, & Hughes, 1992;Schyns & Oliva, 1994). It was found that fine-scale boundary edges (from high spatial frequencies) and coarser scale blobs (from low spatial frequencies) could selectively mediate different categorizations of the same stimuli (e.g., Schyns & Oliva, 1999).…”

Coarse Blobs or Fine Edges? Evidence That Information Diagnosticity Changes the Perception of Complex Visual Stimuli

“…On this third axis, each resolution of the image represents a different spatial scale (in the figure, there are two such spatial scales). studies have dealt with scale-specific recognition of real-world visual displays (though see, e.g., Costen, Parker & Craw, 1994;Hayes, Morrone & Burr, 1986;Parker, Lishman, & Hughes, 1992). In fact, many visual recognition studies (especially those using simple line-drawings of stimuli, e.g., Biederman, 1988;Boyce, Pollatsek, & Rayner, 1989) implicitly assume that important processing mostly occurs at finer resolutions.…”

“…The claim is that a temporal delay between low and high spatial frequency channels (i.e. coarse processing precedes fine processing) would explain the precedence of global information (e.g., Breitmeyer, 1984;Ginsburg, 1986;Parker et al, 1992;Marr, 1982; among many others). However, the psychological literature has often neglected an important difference between global and local processing on one hand, and scale perception on the other.…”