Abstract: The V-system is a complete orthogonal function system which is composed of both
continuous function and functions with discontinuities. In this paper, we propose a new kind of
V-moment functions based on the V-system, and apply them on image classification. Due to the
discontinuity of the basis functions of the V-system, the V-moment functions have distinct advantages in
describing the shapes with a plurality of closed boundaries. When they are applied on feature extraction
for complex shapes, the extracted features are fairly accurate, thus effective image classification
technique can be obtained using the V-moment functions. Experiment of image classification is
conducted on several benchmark databases. The results show that the proposed method has higher
accuracy than Zernike moments, invariant moments and geometric center moments, and it is not
sensitive to noise. Especially, the proposed method presents obvious advantage when it is applied to
classify complex shapes with several closed boundaries.

Key words: image classification, V-system, moment function, V-moments, regional feature