Sensitivity to luminance and chromaticity gradients in a complex scene

Area of Science:

• Computer Vision
• Human Visual Perception
• Computational Neuroscience

Background:

• Image gradients, representing smooth changes in color and luminance, stem from surface properties or illumination effects like shading and shadowing.
• These gradients offer visual cues for inferring factors such as surface orientation relative to light sources.
• Color gradients from mutual illumination (MI) and luminance gradients are crucial for 3D shape-from-shading algorithms, with 3D shape perception influencing MI's effect on color perception.

Purpose of the Study:

• To investigate human sensitivity to changes in color and luminance gradients caused by varying light source positions in complex scenes.
• To determine if observers can discriminate between gradients resulting from different light source locations.
• To understand the role of luminance and chromaticity information in gradient discrimination.

Main Methods:

• Experiment 1: Tested observers' ability to discriminate gradients from different light source positions.
• Experiments 2 and 3: Assessed the reliance on luminance information for gradient detection.
• Experiment 4: Examined the use of spatial distributions of chromaticity and luminance values.

Main Results:

• Observers reliably detected gradient changes when the light source position varied by as little as 4 degrees.
• Sensitivity to gradient changes was primarily based on luminance information.
• A subset of observers utilized the spatial distribution of chromaticity and luminance within gradients for discrimination.

Conclusions:

• Human visual system exhibits high sensitivity to variations in image gradients.
• This sensitivity, particularly to luminance gradients, suggests gradients are vital for recovering 3D shape and scene configuration.
• The findings support the role of image gradients in visual scene understanding and 3D perception.