Effects of predictions robustness and object-based predictions on subjective visual perception.

Area of Science:

• Cognitive Psychology
• Neuroscience
• Computer Vision

Background:

• Contextual associations between objects and scenes aid in predicting environmental features.
• This predictive capability facilitates the perception of visually noisy inputs.
• Prior research indicates predicted objects appear subjectively sharper than unpredicted ones.

Purpose of the Study:

• To investigate if the robustness of context-based predictions modulates the subjective sharpness of objects.
• To explore whether object-based predictions influence the perceived sharpness of scene contexts.
• To understand the reciprocal influences between object and scene predictions in visual perception.

Main Methods:

• Participants performed blur-matching tasks involving images of objects in context and scenes.
• Robustness of predictions was manipulated by altering phase coherence in scene contexts.
• Object predictability was manipulated by using intact versus phase-scrambled objects in scenes.

Main Results:

• Robustly predicted objects were perceived as sharper than less predictable ones when the target object was noisy.
• Predictable scenes were perceived as sharper than unpredictable scenes at equal objective blur levels, specifically when the target scene was noisy.
• Perceptual sharpening was most evident with noisy visual signals and robust predictions.

Conclusions:

• Perceptual sharpening is primarily driven by robust predictions that disambiguate noisy visual information.
• Object-based predictions reciprocally influence the perceived sharpness of scene contexts.
• Visual perception is shaped by the interplay between object-level and scene-level predictions.