Phase-Based Image Symmetry Detection And Bilateral Symmetry Measurement

With the rapid development of information technology, the amount of image information increases greatly. As a result, more and more researchers begin to study the techniques of image retrieval, which are key technical basis of commercially valuable applications. Content-based image processing, in which description of image visual features is a significant part, is very important to the development of economy and society. Symmetry is vital in the description of visual features. Not only is it a crucial feature of objects, but also an important part of HVS(Human Visual System). The detection and measurement for every kind of symmetry is a promising research aspect. Detecting symmetry from complicated images and measuring it is very helpful to pattern recognition and image processing .In this paper, from the characteristics of local phase in the symmetry points of signals, a new method of image symmetry detection is purposed which uses phase information of original image. Moreover, a method of measuring the degree of bilateral symmetry in gray-scale images is put forward which makes full use of the intensity values of object.In this paper, some existing detection and measurement methods for symmetry are discussed, and their performances are analysed.Then, from the analysis of the importance of phase information to the image features and the characteristics of local phase in the symmetry points of ID signals, a new method of symmetry detection is put forward. It uses phase information of original image. Because of phase information's stability and significance, this method is fairly robust, that is, it is invariant to illumination, contrast, and rotation. It can detect mirror symmetry, rotating symmetry and curve symmetry simultaneously. And it doesn't need any preprocessing, such as segmentation. The experiments show the method's effectiveness and robustness.Finally, a method for measuring the degree of bilateral symmetry in gray-scale images is put forward. The method makes use of the intensity values of the object in image. Adopting optimal threshold to segment the original image, it is insensitive to brightness and contrast. The experiments show that the method can correctly indicate the degree of bilateral symmetry of object in every direction.