Related Experiment Videos
Putting the visual system noise back in the picture.
1NASA Ames Research Center, Moffett Field, California 94035.
Summary
This study provides computable expressions for visual system noise in nonlinear models, extending the equivalent-noise concept to analyze stimulus masking and discrimination. These findings offer insights into efficient picture-coding algorithms by simulating visual noise.
Area of Science:
- Visual perception and computational neuroscience.
- Image processing and signal detection theory.
Background:
- Understanding visual system noise is crucial for modeling perception.
- Existing models often simplify internal noise as signal-independent.
Purpose of the Study:
- To derive computable expressions for input-picture-equivalent contrast noise in nonlinear visual models.
- To extend the equivalent-noise concept to signal-dependent internal noise scenarios.
- To analyze the impact of masking stimuli on equivalent noise structure.
Main Methods:
- Developed computable expressions for equivalent contrast noise within a locally linear subclass of nonlinear models.
- Allowed for signal-dependent internal model noise.
- Extended the equivalent-noise concept to models explaining suprathreshold masking and discrimination.
Main Results:
- Provided expressions for equivalent noise that depend on the masking stimulus.
- Demonstrated that noise structure is influenced by masker representation at performance-limiting noise generation levels.
- Showed applicability to less-than-full-rank transformations.
Conclusions:
- Computable expressions for visual system noise can be derived for complex nonlinear models.
- Equivalent noise is stimulus-dependent and linked to internal noise generation.
- Simulating visual noise in images can inform efficient picture-coding strategies.