Purpose: Because most man-made objects and animals are symmetric, detecting symmetric objects is important for humans. Indeed, human observers have little difficulty in determining whether a given object is symmetric or not, regardless of the viewing direction. Prior research concentrated on the case where the retinal image itself was symmetric. However the retinal image of a symmetric object is symmetric only for a small set of viewing directions. We tested perception of planar (2D) symmetric objects when the objects were slanted in depth and the retinal image itself was not symmetric. Experiments: In the first experiment, we tested symmetry detection with dotted patterns and polygons. Symmetry can be detected reliably in the case of polygons, but not dotted patterns. In the second experiment, we tested the role of image features representing the symmetry of a 2D figure (orientation of the projected symmetry axis and symmetry lines) vs. that representing the 3D viewing direction (orientation of the axis of rotation). Symmetry detection is more reliable when the projected symmetry axis or lines are known to the subject, but not when the axis of rotation is known. Finally, we tested symmetry detection with orthographic and perspective images. Symmetry detection with orthographic images is slightly better than that with perspective images. Based on these results, we will propose a computational model of symmetry detection. This model operates on shapes of contours, rather than on corresponding vertices. Generalization of these results to the case of images of volumetric (3D) symmetric objects will be discussed.