Abstract: A novel shape image retrieval algorithms based on compensation mechanism under miminum circumscribed circle was proposed in this paper, which not only extract the features of the target area, but also compensate the background characteristics of the object under miminum circumscribed circle. Extracting the Hu invariant moments and Zernike moment invariants,commonly used in the description of the regional characteristics,as image features. Moment invariants have the unchanged characteristics of translation, scaling and rotating. Similarity is given by the Euclidean distance between two images' normalized moment vectors. hhe approch is implemented simply, and compensates the human eye's visual experience effectively. A large number of experiments shown that the approach can achieve better accuracy and recall rate than that based on an object area only.

Key words: Background compensation, Hu momnets, Zernike moments, Shape, Image retrieval