Abstract
Humans recognize objects across different viewpoints with remarkable ease. Yet much remains unknown about how the visual system accomplishes this incredible feat of recognition. One intriguing proposal is that object recognition is so efficient because vision relies on a representation of the medial axis, or the so-called shape skeleton (Blum, 1973; Firestone & Scholl, 2013). Support for this proposal comes from research demonstrating that humans have an attentional bias for the medial axis when tasked with tapping anywhere they would like inside a shape (Firestone & Scholl, 2014). However, there exists very little behavioral evidence that further confirms the existence of a medial axis representation. The current study provides converging evidence for the medial axis model by demonstrating its existence in a novel context. Using a visual search paradigm, we tested whether target detection within a rectangle was faster along the medial axis. Participants were tasked with detecting a darker square among an array of distractor squares (132 total) arranged in a rectangle. For each participant, the shade of the target was titrated using a color sensitivity task that also served as training. Reaction times (RTs) were recorded for all targets within the rectangle. Preliminary results indicate that participants are faster at detecting targets that are located along the medial axis of the rectangle in comparison to other (i.e., non-medial axis) points in the array (p < .05). Moreover, RTs decreased as a function of target distance to the medial axis; that is, the closer the target was to the medial axis, the faster participants' detected it (p < .001). These findings provide further support for a medial axis model of object recognition and expand on previous studies that have shown enhanced contrast sensitivity of Gabor filters along the medial axis of a shape (Kovács & Julez, 1994; 1998).
Meeting abstract presented at VSS 2016