Detection of partial symmetry using correlation with rotated-reflected images

摘要

Symmetry is one of the important structural properties of a figure both in visual psychology and in computer vision. Symmetrical properties of planar two-dimensional figures are considered and a new method for the extraction of rotational symmetry and reflectional symmetry is presented. A figure is said to have symmetry if it is invariant with some congruent transformation which consists of translation, rotation and reflection. The approach is to obtain the congruent transformations which make the transformed image a good match with the original image. The directional correlation of edge features is used to evaluate how well the original and transformed images match. This method does not require any segmentation nor knowledge of the centroid position, and the symmetrical properties of partially symmetric figures can also be detected. The approach is described and this method is applied to synthetic and real images.