Membership matching score for invariant image recognition

This paper describes the use of membership matching score (MMS) to solve the recognition errors in invariant image recognition. The method works on boundary normalization of images, which to move the starting point back on the semi major axis of the ellipses. The set of maximum and minimum values are computed as the boundaries for the groups of the images. Then, the voting score of the membership of unknown boundary function in the inner boundary set are used as the indicator to measure the similarity level. The proposed method in this research is effective to solve the problem of invariant images. The results shown that, the elliptic Fourier descriptors and MMS are an efficient representation which can provide for reliable recognition.