Symmetry Detection Based on Phase Information and Principal Component Analysis

Symmetry detection plays an important role in image analysis and pattern recognition. Based on phase symmetry and principal component analysis, a new image symmetry detection method is proposed. Firstly, phase symmetry is computed on different scales and orientations. On each orientation, the phase symmetry values of different scales are merged together. And then the main feature of different orientations is extracted by using principal component analysis. Finally, the results of symmetry detection can be obtained by using non-maximal suppression and adaptive hysteresis thresholding. The experiments show that the proposed method can be applied directly to original images and it does not need segmentation or any preprocessing. And it is insensitive to rotation, brightness and contrast. It also can detect mirror symmetry, rotational symmetry and curve symmetry at the same time, whether bright or dark objects.