Image recognition by combined invariants of legendre moment

The moment invariants are usually efficient tools in image processing, pattern recognition and computer vision, especially when images to be processed are degraded by combined blur and affine transformation. However, most of the existing combined moment invariants are based on nonorthogonal moments, which can result in the information redundancy and become sensitive to noise. Due to the better performance of orthogonal moment, this paper presents a new set of orthogonal Legendre moments affine invariants and then constructs Legendre combined invariants with respect to blur and affine transformation by combining blur invariants and affine invariants. The experimental results are provided in order to validate the method and to compare its performance with other combined invariants.