Fast and accurate computation of orthogonal moments for texture analysis

摘要: Abstract In this work we describe a fast and stable algorithm for the computation of orthogonal moments an image. Indeed, are characterized by high discriminative power, but some their possible formulations large computational complexity, which limits real-time application. This paper describes in detail approach based on recurrence relations, proposes optimized Matlab implementation corresponding procedure, aiming to solve above limitations put at community’s disposal efficient easy use software. our experiments evaluate effectiveness formulation, as well its performance reconstruction task, comparison closed form representation, often used literature. The results show sensible reduction together with greater accuracy reconstruction. order assess compare computed texture analysis, perform classification six well-known databases images. Again, formulation performs better than representation. More importantly, if from GLCM image using proposed outperform situations most diffused state-of-the-art descriptors classification.