Scientists from York’s Centre for Vision Research measured how quickly and accurately participants could identify images of animals on a computer screen. They found that the key cues to animal detection are shape and texture, while the influence of colour and luminance are minimal.

“It turns out that although a computer algorithm can use colours to discriminate animal from non-animal images fairly reliably, humans cannot,” says study lead author James Elder, a member of York’s Centre for Vision Research, and Associate Professor of Computer Science & Engineering and Psychology. “If, on the other hand, humans are presented with an image that contains only the shapes in an image, they can very rapidly determine whether an animal is present.”

Prior work on rapid identification of natural scenes has suggested that this may be based largely on texture features.

“Our results show that texture can also be used to detect animals, but processing of texture features is delayed relative to contour shape cues,” Elder says.

Prior to this research, little had been known about the mechanisms underlying these image-processing abilities, including how such cues are processed over time. Previous studies have shown that the neural signals necessary for recognizing an image appear as early as 150 milliseconds after we see it.

In one experiment, participants were shown a series of full colour, monochrome, and paint-by-number images (created from an average of colours), along with contour images for which only outlines were displayed.

“It appears that the visual system relies upon a rapidly-extracted contour representation of a scene,” says Elder. “Colour and luminance may have some influence late in the course of stimulus processing, but its influence is fairly minor relative to that of shape and texture,” he says.

Elder points out the importance of this research in understanding how our brains evolve and adapt to our surroundings.

“The task of rapidly detecting and identifying another nearby animal has no doubt helped to drive the evolution of our visuals systems,” he says. “Taken as a whole, this research furthers our understanding of the neural mechanisms that have evolved to solve practical visual tasks.”

The paper, “Cue dynamics underlying rapid detection of animals in natural scenes,” was published online in July 2009, in the Journal of Vision. It is co-authored by Ljiljana Velisavljevic, now at the University of Toronto.

York University is the leading interdisciplinary research and teaching university in Canada. York offers a modern, academic experience at the undergraduate and graduate level in Toronto, Canada’s most international city. The third largest university in the country, York is host to a dynamic academic community of 50,000 students and 7,000 faculty and staff, as well as 200,000 alumni worldwide. York’s 10 Faculties and 28 research centres conduct ambitious, groundbreaking research that is interdisciplinary, cutting across traditional academic boundaries. This distinctive and collaborative approach is preparing students for the future and bringing fresh insights and solutions to real-world challenges. York University is an autonomous, not-for-profit corporation.

-30-

Media Contact:

Melissa Hughes, Media Relations, York University, 416 736 2100 x22097, mehughes@yorku.ca