The roles of spatial frequency in category‐level visual search of real‐world scenes

Abstract

Visual object recognition appears to be effortless and virtually instantaneous for humans. A number of previous investigations have suggested that visual perception follows a coarse‐to‐fine processing sequence in which low‐spatial‐frequency (LSF) information provides a coarse representation of an object for an initial guess and guides the fine processing of the object based on high‐spatial‐frequency (HSF) information. A category‐level visual search of real‐world scenes shares many aspects of processing mechanisms with object recognition. However, a key difference between them is the target template in visual search that is generally known in advance. In the present study, we investigated the role of spatial frequency information in the category‐level visual search of real‐world scenes with three behavioral experiments. In Experiment 1, we showed that the representation of the target template, which is held in short‐term memory, had biased attention more towards LSF information than HSF information in the probe trials. In Experiments 2 and 3, we further demonstrated that the near‐threshold LSF information could implicitly facilitate object recognition compared with HSF information, thereby providing supporting evidence that the LSF information of the target template implicitly facilitates target recognition and search performance. These findings support the role of the coarse‐to‐fine principle in rapid visual search performance of real‐world scenes.

Perception of attractive and unattractive face groups is driven by distinct spatial frequencies

“…Evidence from behavioral experiments and neuroscience studies suggests that face perception is hierarchical, and involves multiple brain circuits (Bruce & Young, 1986; Calder & Young, 2005; Haxby, Hoffman, & Gobbini, 2000; Kanwisher, McDermott, & Chun, 1997; Maurer, Le Grand, & Mondloch, 2002; Xu, Dayan, Lipkin, & Qian, 2008). Each circuit may be specialized to process specific aspects of facial characteristics, including spatial frequencies (SFs; Badcock, Whitworth, Badcock, & Lovegrove, 1990; DeValois, Albrecht, & Thorell, 1982; Goffaux, Gauthier, & Rossion, 2003; Goffaux & Rossion, 2006; Halit, de Haan, Schyns, & Johnson, 2006; Jeantet et al, 2019; Zhang & Li, 2019). For example, the amygdala, an important subcortical region for emotion perception, favors low spatial frequency (LSF) information (Said, Baron, & Todorov, 2009); while the fusiform face area, one of the core regions for face perception, receives visual inputs at various SFs (Gauthier, Curby, Skudlarski, & Epstein, 2005).…”

Grounding Context in Embodied Cognitive Robotics

“…Visual recognition of scenes is a fast, automatic, and reliable process (Oliva, 2005 ; Greene and Oliva, 2009 ; Lowe et al, 2018 ; Kaiser et al, 2019 ). Thorpe et al ( 1996 ) have reported that complex natural scenes can be categorized under 150 ms. To explain this phenomenon, theories of visual perception have suggested a mode of processing based on specific spatial frequencies that would convey different information about the appearance of a stimulus (Kauffmann et al, 2015 ; Zhang and Li, 2019 ; Aghajari et al, 2020 ). High spatial frequencies (HSFs) represent abrupt spatial changes in visual information such as edges and correspond to configuration information and fine detail.…”