Perceived intrinsic 3D shape of faces is robust to changes in lighting direction, image rotation and polarity inversion.

Abstract

Face recognition from 2D images is influenced by various factors, including lighting conditions, viewing direction, rotation, and polarity inversion. It has been proposed that these techniques affect face recognition by distorting shape from shading. This study investigates the perception of 3D face shape in 2D images using a gauge figure task. Two experiments were conducted where participants adjusted a gauge figure across multiple locations within a 3D image to assess its surface structure. We manipulated face orientation, lighting direction, and polarity inversion (exp 2). While these manipulations resulted in variations from the true surface structure, they could be explained by an affine transformation. This suggests that the perception of the intrinsic 3D shape of faces is stable across these image manipulation techniques. The effects of viewing conditions on face recognition may thus be better interpreted through their influence on the perception of material properties such as pigmentation, or on information closer to the level of the retinal image itself.