The perception of 3D shape from texture based on directional width gradients.

Area of Science:

• Computer vision
• Computational geometry
• Human perception

Background:

• Estimating 3D surface shapes from 2D images is a fundamental problem in computer vision.
• Anisotropic textures, where properties vary with direction, pose unique challenges for shape estimation.
• Existing methods often struggle with surfaces exhibiting non-uniform textural properties.

Purpose of the Study:

• To develop and validate a computational analysis for estimating 3D surface shapes from anisotropic textures under negligible perspective.
• To investigate the impact of texture homogeneity on perceived surface shape.
• To test the model's predictions using psychophysical experiments.

Main Methods:

• A novel computational analysis was developed to model 3D shape from anisotropic textures.
• The analysis assumes texture homogeneity and predicts shape distortions when violated.
• Two psychophysical experiments were conducted to test these predictions.

Main Results:

• Observer judgments of surface shape were consistent with the computational model's predictions.
• The model successfully predicted how violations in texture homogeneity influence perceived 3D shape.
• Ordinal shape judgments aligned with the theoretical framework.

Conclusions:

• The developed computational analysis provides a robust method for 3D shape estimation of curved surfaces with anisotropic textures.
• Texture homogeneity is a critical factor influencing human perception of surface shape.
• The study highlights the limitations of the current model for surfaces viewed with significant perspective, suggesting avenues for future research.