Gravitational effects of scene information in object localization

Area of Science:

• Visual perception
• Cognitive neuroscience
• Computational vision

Background:

• Object location perception is crucial for environmental interaction.
• Apparent object position deviates from retinotopic location due to contextual influences.
• The impact of complex natural scene visual information on perceived object location is not well understood.

Purpose of the Study:

• To investigate how local image statistics in natural scenes influence the reported location of visual targets.
• To determine if image features like luminance, edges, object boundaries, and saliency bias perceived location.
• To explore the relationship between these image statistics and saccade errors.

Main Methods:

• Observers reported the location of a brief target superimposed on natural scene images.
• The influence of local image statistics (luminance, edges, object boundaries, saliency) was quantified by comparing physical and reported target centers.
• Difference scores were analyzed against a null distribution to isolate image-specific effects and account for response biases.

Main Results:

• Reported locations were significantly biased towards darker regions, luminance edges, object boundaries, and areas of high saliency.
• These image statistics showed relatively low shared variance, suggesting distinct influences.
• Similar image statistics influenced saccade errors, persisting even when high-level scene processing was disrupted.

Conclusions:

• Natural images exhibit landmark effects, where perceived feature locations are biased by informative low- and high-level visual content.
• These findings highlight the role of image statistics in shaping spatial awareness within complex visual environments.
• The observed biases in location reporting and saccade errors underscore the integral role of scene context in visual guidance.