Towards a unified theory of 3D shape perception

Objectives. Unifying theoretical analysis of visual 3D shape cues, including texture, shading and highlights. Experiments test whether we experience illusions of shape predicted by the model. Methods. Model uses the population response of filters tuned to different image orientations (cf. V1 cells). Subjects viewed rendered images of 3D objects and adjusted ‘gauge figure’ probes to report perceived surface orientation. Results: Model correctly predicts both success and failures of human 3D shape estimation across variations in texture, lighting and surface reflectance properties. Conclusions: For the early stages of 3D shape estimation, seemingly different cues may have more in common than previously believed