Visual search in noise: Revealing the influence of structural cues by gaze-contingent classification image analysis

Abstract

Visual search experiments have usually involved the detection of a salient target in the presence of distracters against a blank background. In such high signal-to-noise scenarios, observers have been shown to use visual cues such as color, size, and shape of the target to program their saccades during visual search. The degree to which these features affect search performance is usually measured using reaction times and detection accuracy. We asked whether human observers are able to use target features to succeed in visual search tasks in stimuli with very low signal-to-noise ratios. Using the classification image analysis technique, we investigated whether observers used structural cues to direct their fixations as they searched for simple geometric targets embedded at very low signal-to-noise ratios in noise stimuli that had the spectral characteristics of natural images. By analyzing properties of the noise stimulus at observers' fixations, we were able to reveal idiosyncratic, target-dependent features used by observers in our visual search task. We demonstrate that even in very noisy displays, observers do not search randomly, but in many cases they deploy their fixations to regions in the stimulus that resemble some aspect of the target in their local image features.

Visual orienting in dynamic broadband (1/f) noise sequences

“…For instance, using the method of classification images, Rajashekar, Bovik, and Cormack (2006) found that saccadic target selection was based on structural cues, despite the noisiness of the 1/f displays. Tavassoli, van der Linde, Bovik, and Cormack (2007) continued that line of research, but additionally proposed a variant of the classification image paradigm, with which one can obtain useful classification images in as few as 200 trials.…”

“…The noise movie was created from a two-dimensional 1/f image (Figure 1)-the same type of image as has been used in other studies (e.g., Geisler et al, 2006;Rajashekar et al, 2006). In those studies, the image was presented statically, but in our present study, it was used as the source for a one-dimensional movie by iterating row-wise through the source image: Each row was stretched vertically in space and presented as a frame ( Figure 2); the movie thus appeared as a flickering bar code.…”

“…The method is typically applied to image patches of "false alarms" in experimental tasks containing the four response outcomeshits, misses, false alarms, and correct rejections-or by linear combination over classification images for all response outcomes. The classification images typically reveal aspects of stimulus representation and recognition mechanisms when Gaussian noise is used, but recently the methodology has been applied to 1/f noise (see, e.g., Geisler et al, 2006;Rajashekar et al, 2006). In our present study, saccadic landing positions can be loosely regarded as false alarms, because no actual target has been placed at those locations, and the classification image may therefore contain hints about the presence of a possible trigger feature for saccadic target selection in 1/f displays.…”

Similar Neural Representations of the Target for Saccades and Perception during Search

“…We used classification image analysis (Ahumada and Lovell, 1971;Gold et al, 2000;Eckstein and Ahumada, 2002;Neri and Heeger, 2002;Levi and Klein, 2003;Rajashekar et al, 2006), which allowed us to measure and compare the behaviorally defined, spatial receptive fields of the visual mechanisms responsible for saccadic and perceptual decisions. With this method, the entire brain is viewed as a single system in which input is the stimulus and output is either the saccadic or perceptual decision.…”

The SNCD as a Metrics for Image Quality Assessment

“…[12] propose how to quantify lost image information and explore a relationship between image information and image quality. The authors of [13] investigated whether observers used structural cues to direct their fixation as they searched for simple embedded geometric targets at very low signal-to-noise ratios; the authors demonstrated that even in case of very noisy displays, observers do not search randomly, but in many cases they deploy their fixation to stimulus regions that resemble some aspect of the target in their local image features. [2] show an evaluation of different recent full reference image quality assessment methods, where they performed a subjective evaluation.…”