How Different Spatial-Frequency Components Contribute to Visual Information Acquisition.

Abstract

We test 3 theories of global and local scene information acquisition, defining global and local in terms of spatial frequencies. By independence theories, high-and low-spatial-frequency information are acquired over the same time course and combine additively. By global-precedence theories, global information acquisition precedes local information acquisition, but they combine additively. By interactive theories, global information also affects local-information acquisition rate. We report 2 digit-recall experiments. In the 1st, we confirmed independence theories. In the 2nd, we disconfirmed both independence theories and interactive theories, leaving global-precedence theories as the remaining alternative. We show that a specific global-precedence theory quantitatively accounted for Experiments 1-2 data as well as for past data. We discuss how their spatial-frequency definition of spatial scale comports with definitions used by others, and we consider the suggestion by P. G. Schyns and colleagues (e.g., D. J. Morrison & Schyns, 2001) that the visual system may act flexibly rather than rigidly in its use of spatial scales.Since the time of Neisser's (1967) classic Cognitive Psychology, visual scenes have been conceptualized as being decomposable into global and local information: Global information corresponds to overall scene structure, whereas local information corresponds to fine details (see Morrison & Schyns, 2001, pp. 454 -456, for a summary). This article is about the relations between global information and local information-the relative time courses over which they are acquired and the means by which they combine into an overall perception of the scene. Spatial-Frequency InformationGlobal and local information can be operationalized in many ways. One of them is in terms of spatial frequencies. From the visual system's perspective, the world is made up of different spatial frequencies. An example is provided in Figure 1, the top of which shows a typical real-world scene. The two bottom panels show the same picture filtered to pass only the low spatial frequencies (LSFs; bottom left) or high spatial frequencies (HSFs; bottom right). It is evident that the two spatial-frequency components carry different kinds of information that accord well with the general conceptualization of global information and local information: The bottom left Figure-1 picture carries a global representation of the scene, whereas the bottom right picture conveys information about edges and details, such as the writing on the dollar bill. There is indeed ample evidence that the visual system decomposes the visual scene into separate spatial-frequency components (Blakemore & Campbell, 1969;Campbell & Robson, 1968;De Valois & De Valois, 1980, 1988Graham, 1989;Olzak & Thomas, 1986). Accordingly, a correspondence between global-local on the one hand and LSF-HSF on the other has been hypothesized by numerous investigators (e.g., Hughes, Nozawa, & Ketterle, 1996;Schyns & Oliva, 1994), and we shall use it in the remainder of this article....

Shining in the Center: Central Gaze Cascade Effect on Product Choice

“…Participants' eye movements were recorded at a rate of 50 samples per second; therefore, each time bin covered 20 milliseconds. Detection and categorization of visual stimuli is accurate in fixations as brief at 17 milliseconds (Grill-Spector and Kanwisher 2005), and overall categorization accuracy is ensured between 20-100 milliseconds (Loftus and Harley 2004;Loftus and McLean 1999). Therefore the use of a 20-millisecond time bin allows accounting for each fixation in which information would have been extracted.…”

The singular nature of auditory and visual scene analysis in autism

“…The visual system analyses visual images using spatial-frequency channels. Low spatial frequencies convey the global pattern of the visual scene whereas high spatial frequencies carry the fine structure of the scene [40]. It should be also noted that functionally similar subprocesses in the two modalities are probably implemented at anatomically different levels.…”

Image Quality Assessment Using Spatial Frequency Component

“…The image quality can be quantified separately in different spatial frequencies, and then the results from different spatial frequencies are fused to the final score. Here we consider the decomposition into LSF and HSF [18]. As shown in Figure 2, the left column shows the reference image and the distortion images with luminance shift and add-in noise respectively.…”

“…the SFCGL, which considers both the global and local information distortion of the image. As shown in Fig.3, it consists of four steps: (1) spatial frequency decomposition of the input image into LFS and HFS [18]; (2) measurement of the global information distortion in LFS, where the rectified mean absolute difference is used; (3) measurement of the local information distortion in HFS, where the SSIM metric is adopted; and (4) fusion of the measurements using the abruptness measures the residual images of LFS and HFS to predict the observer's judgment.…”

“…On the other hand, Fig.1 (d)~(f) give a failure example of the VSNR, which focus on the global spatial frequency component. It can be seen that both (e) and (f) in Fig.1 Actually, according to the global-to-local theory (GTL) [18][19][20][21][22][23], both HSF and LSF are very important in IQA, although they influence the visual effect of image in different ways. Inspired by the above fact, we propose a novel image quality metric (namely Spatial Frequency Component with Global and Local information SFCGL) that considers both the global and local distortions existing in different spatial frequency components of the image.…”