Illusions are in the eye, not the mind
IMAGE: THE BAR IN THE MIDDLE OF THIS FIGURE IS ALL ONE GREY LEVEL, BUT IT APPEARS LIGHTER ON THE LEFT AND DARKER ON THE RIGHT DUE TO THE GRADIENT IN THE BACKGROUND. THIS IS CALLED SIMULTANEOUS CONTRAST, WHERE DARK SURROUNDS MAKE TARGETS APPEAR LIGHTER, AND VICE-VERSA. view more
CREDIT: JOLYON TROSCIANKO
Numerous visual illusions are caused by limits in the way our eyes and visual neurones work – rather than more complex psychological processes, new research shows.
Researchers examined illusions in which an object’s surroundings affect the way we see its colour or pattern.
Scientists and philosophers have long debated whether these illusions are caused by neural processing in the eye and low-level visual centres in the brain, or involve higher-level mental processes such as context and prior knowledge.
In the new study Dr Jolyon Troscianko, from the University of Exeter, co-developed a model that suggests simple limits to neural responses – not deeper psychological processes – explain these illusions.
“Our eyes send messages to the brain by making neurones fire faster or slower,” said Dr Troscianko, from the Centre for Ecology and Conservation on Exeter’s Penryn Campus in Cornwall.
“However, there’s a limit to how quickly they can fire, and previous research hasn’t considered how the limit might affect the ways we see colour.”
The model combines this “limited bandwidth” with information on how humans perceive patterns at different scales, together with an assumption that our vision performs best when we are looking at natural scenes.
The model was developed by researchers from the Universities of Exeter and Sussex to predict how animals see colour, but it was also found to correctly predict many visual illusions seen by humans.
“This throws into the air a lot of long-held assumptions about how visual illusions work,” Dr Troscianko said.
He said the findings also shed light on the popularity of high-definition televisions.
“Modern high dynamic range televisions create bright white regions that are over 10,000 times brighter than their darkest black, approaching the contrast levels of natural scenes,” Dr Troscianko added.
“How our eyes and brains can handle this contrast is a puzzle because tests show that the highest contrasts we humans can see at a single spatial scale is around 200:1.
“Even more confusingly, the neurones connecting our eyes to our brains can only handle contrasts of about 10:1.
“Our model shows how neurones with such limited contrast bandwidth can combine their signals to allow us to see these enormous contrasts, but the information is ‘compressed’ – resulting in visual illusions.
“The model shows how our neurones are precisely evolved to use of every bit of capacity.
"For example, some neurones are sensitive to very tiny differences in grey levels at medium-sized scales, but are easily overwhelmed by high contrasts.
“Meanwhile, neurones coding for contrasts at larger or smaller scales are much less sensitive, but can work over a much wider range of contrasts, giving deep black-and-white differences.
“Ultimately this shows how a system with a severely limited neural bandwidth and sensitivity can perceive contrasts larger than 10,000:1.”
The paper, published in the journal PLOS Computational Biology, is entitled: “A model of colour appearance based on efficient coding of natural images.”
The two grey bars in the middle of this figure are the same grey, but the one on the left (surrounded by more black bars) appears darker. This is the opposite of the simultaneous contrast example above, because darker surrounds now make the target look darker.
Both cubes have what appear to be yellow and blue tiles on their top surfaces. However, the ones that look yellow on the left are in fact a grey colour that is identical to the blue tiles on the right. Our model can help explain how objects appear to be the same colour even when the light changes, and why in illusions such grey looks colourful.
CREDIT
Jolyon Troscianko
JOURNAL
PLoS Computational Biology
ARTICLE TITLE
A model of colour appearance based on efficient coding of natural images
ARTICLE PUBLICATION DATE
15-Jun-2023
No comments:
Post a Comment