Local Energy Based Symmetry Detection Algorithm

Most of the existing symmetry detectors are based on luminance or gradient information. Upon analyzing the relationship between symmetry and phase, we propose a new symmetry detection algorithm based on local energy. The symmetry detection is converted to energy analysis by the approach. In frequency domain, we filter the original image by log Gabor wavelet, and generate those symmetric points of the objects, which have the maximal local energy in various scales and keep their phases in certain special phases and congruent. The feasibility analysis, local Energy based Symmetry Detection definition, and rationality demonstrations establish a theoretic foundation for this algorithm. Experimental results show that it can overcome the limitations of the existing algorithms in demanding segmentation prior to any symmetry detection. This algorithm can be applied directly to original images and it needs no segmentation and any other preprocessing. It is invariant to image rotation and can detect various types of symmetries at the same time.