Related Experiment Videos
A contrast- and luminance-driven multiscale network model of brightness perception
L Pessoa1, E Mingolla, H Neumann
1Department of Cognitive and Neural Systems, Boston University, MA 02215, USA.
Vision Research
|August 1, 1995
Summary
A new neural network model explains brightness perception, accurately predicting phenomena like Mach bands and contrast effects. This model advances understanding of visual processing by integrating boundary and feature signals.
Area of Science:
- Vision Science
- Computational Neuroscience
- Psychophysics
Background:
- Brightness perception is complex, involving interactions between luminance, contrast, and spatial context.
- Existing models struggle to explain a wide range of perceptual phenomena comprehensively.
- The concept of brightness 'anchoring' remains a challenge in visual modeling.
Purpose of the Study:
- To develop a novel neural network model of brightness perception.
- To account for diverse visual data, including Mach bands and contrast effects.
- To provide a new interpretation of feature signals and brightness anchoring.
Main Methods:
- Developed a neural network model incorporating boundary and feature signal computations.
- Modeled sensitivity to luminance steps and gradients.
- Explicitly represented contrast-driven and luminance-driven information.
Main Results:
- The model successfully predicts phenomena such as Mach bands and missing fundamental effects.
- It accounts for non-linear contrast effects with sinusoidal waveforms.
- Demonstrated competence in simulating various brightness perception data.
Conclusions:
- The proposed neural network model offers a unified explanation for diverse brightness perception phenomena.
- The model's approach to feature signals provides insight into brightness anchoring.
- This work advances computational models of visual perception.