Improved Moment Invariants for Invariant Image Representation

This paper proposes improved moment invariants for representing images. These features are invariant to orientation, size and translation. These new features compute moments from a point shifted to a distance from the image centroid. By doing so, the new moments show improved ability to represent symmetrical and noisy images. Both these problems of symmetry and noise are common when regular moment invariants are used. The new reference centre is selected such that the invariant properties like translation, scaling and rotation are maintained. In this paper, we show that these new proposed moments solve the symmetrical problem and are more robust to noise corruption as compared to two different types of regular moment functions derived by Hu [9]. Extensive experimental study using a neural network classification scheme with these moments as inputs are conducted to verify the proposed method.