Image Retrieval Using Ellipse Shape Feature with Particle Swarm Optimization

The paper presents a method of image retrieval using ellipse shape feature and particle swarm optimization. Based on Fourier descriptor (FD), this method decomposes the contour of image into a series of ellipses. An ellipse can be simply represented with semi-major axis, semi-minor axis, and orient angle. Because the distinction of different ellipses´ parameters is remarkable, the weights of parameters in similarity computation between two images can be finely tuned by particle swarm optimization. Experimental results on a typhoon cloud image database indicate that the proposed method is capable of effectively retrieving images and achieving better results than the ordinary FD retrieval method and other conventional retrieval methods.