Researchers have developed a computational model that can predict colour appearance by incorporating the principle of efficient coding. The model takes into account the influence of an object’s surroundings on its colour, brightness, and pattern, and suggests that many visual phenomena and illusions can be attributed to simple mechanisms evolved for efficient coding of natural images. The findings, published in the journal PLOS Computational Biology, have important implications for understanding the vision of humans and other animals.
An object’s colour and brightness are not solely determined by its own surface properties but are also influenced by the colours and patterns of its surroundings. Various visual phenomena and illusions have been discovered that demonstrate the dramatic effects of these influences. While explanations for these phenomena have ranged from low-level neural mechanisms to high-level processes that incorporate contextual information, quantitative models of colour appearance have struggled to account for many of these phenomena.
To address this gap, researchers aimed to develop a model that could predict colour appearance based on the principle of coding efficiency. The model assumes that images are encoded by noisy spatio-chromatic filters at different spatial scales, which can be either circularly symmetrical or oriented. The filters have specific thresholds and dynamic ranges that determine their responses to different stimuli. The model also incorporates contrast sensitivity functions that account for variations in contrast detection thresholds with spatial scale.
The researchers tested the model’s performance by comparing it to human behavioural performance in psychophysics experiments and primate retinal ganglion responses. The results showed that the model accurately predicted human and primate responses, suggesting that it fits well with the underlying mechanisms of colour perception.
Furthermore, the researchers systematically tested the model’s ability to predict over 50 brightness and colour phenomena, including various illusions. The model demonstrated almost complete success in predicting the direction of these phenomena, providing further evidence for its effectiveness.
The model’s simplicity and lack of free parameters are notable strengths. It can qualitatively predict a wide range of perceptual phenomena without the need for complex mechanisms or top-down effects. Instead, the model suggests that many aspects of colour appearance can be attributed to mechanisms evolved for efficient coding of natural images.
The findings challenge previous explanations that rely on high-level interpretations or multiple sources of sensory evidence and prior knowledge. The model’s feed-forward architecture explains why visual phenomena appear without delay and eliminates the need for feedback loops or adaptations.
While the model accurately predicts many visual phenomena, it does have some limitations. It struggles to predict certain illusions, such as illusory spots and bars in the Hermann grid and Poggendorff illusions, and shows comparatively weak performance in specific cases. However, these limitations can be addressed by adjusting the model accordingly.
The spatiochromatic bandwidth limited (SBL) model offers a promising approach to understanding colour appearance based on efficient coding principles. The researchers believe that further investigations should focus on quantitative testing of the model and exploring its performance in various tasks. Additionally, they suggest that future work should measure the bandwidth of chromatic channels and investigate the potential role of single-opponent and double-opponent pathways in colour perception.
The development of this model provides valuable insights into the mechanisms underlying colour appearance and offers new avenues for research in the field of vision science. By uncovering the role of efficient coding in colour perception, scientists can gain a better understanding of how humans and animals perceive and interpret the visual world.
