Similarity measure based on membership function

A novel shape similarity measure based on membership function is proposed in this work to make the measure more consistent with human perception and improve the matching accuracy. The proposed method commences with extraction of feature vectors of shapes in training shape set, then a fuzzy set over each eigenvalue space is defined. The membership function of the fuzzy set is defined and acts as a weight in similarity measure. The performance of the proposed method is compared with two other methods. Our experiment results show that the accuracy of matching has been improved notably by our method. We also studied the characteristics of shape descriptor and L_p distance we used and determined a proper scale of them.