3D image analysis by separable discrete orthogonal moments based on Krawtchouk and Tchebichef polynomials. (November 2017)

• Record Type: Journal Article
• Authors: Batioua, Imad; Benouini, Rachid; Zenkouar, Khalid; Zahi, Azeddine; Hakim, El Fadili
• Is Part Of: Pattern recognition. Volume 71(2017:Nov.)
• Page Start: 264
• Page End: 277
• Publication Date: 2017-11
• Subjects: 3D image analysis -- Separable moments -- Moment invariants -- Numerical stability -- Object classification -- Multivariate discrete orthogonal polynomials -- Krawtchouk moments -- Tchebichef moments; TKKM Tchebichef-Krawtchouk-Krawtchouk Moments -- TTKM Tchebichef-Tchebichef-Krawtchouk Moments -- TTTM Tchebichef-Tchebichef-Tchebichef Moments -- KKKM Krawtchouk-Krawtchouk-Krawtchouk Moments -- TKKMI Tchebichef-Krawtchouk-Krawtchouk Moment Invariants -- TTKMI Tchebichef-Tchebichef-Krawtchouk Moment Invariants -- TTTMI Tchebichef-Tchebichef-Tchebichef Moment Invariants -- KKKMI Krawtchouk-Krawtchouk-Krawtchouk Moment Invariants -- ROI Region Of Interest -- MSE Mean Squared Error; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2017.06.013
• Languages: English
• ISSNs: 0031-3203

Abstract

Highlights: This paper introduces new sets of separable discrete moments for 3D image analysis. This paper provides the process for deriving 3D moment invariants (scale, translation, rotation). Numerical experiments are performed to demonstrate its validity and superiority. Abstract: In this paper, we introduce new sets of separable discrete moments for 3D image analysis, named: TKKM (Tchebichef-Krawtchouk-Krawtchouk Moments) and TTKM (Tchebichef-Tchebichef-Krawtchouk Moments). Firstly, we present a detailed comparative study between the proposed separable 3D moments and the classical ones in terms of global feature extraction capability under noisy and noise-free conditions. Also, their local feature extraction ability is examined. Secondly, our study investigates the ability of the proposed separable 3D moments in pattern recognition. For this, new sets of separable 3D discrete moment invariants are introduced. The proposed rotation, scaling and translation 3D moment invariants have been rigorously tested under different sets of mixed transforms. The obtained results show that the representation capability, in comparison with traditional Krawtchouk and Tchebichef moments, has been significantly improved by using the new proposed 3D separable moments and can be highly useful in the field of 3D image analysis.