Note on the Construction of Structure Tensor

2025-10-27

AI Summary

This work addresses key limitations of the Bigun–Granlund (1987) and Granlund–Knutsson (1995) structural tensor formulations—namely, non-semidefiniteness of the resulting tensors, ambiguous physical interpretation of eigenvalues, reliance on orthogonal filters, and dependence on polar-coordinate spectral partitioning. We propose a unified modeling framework based on total least squares (TLS) linear fitting of the power spectrum. By averaging outer products of gradient responses and weighting with tuned frequency vectors, our method eliminates redundant correction terms and rigorously ensures tensor semidefiniteness. It supports multi-directional fitting and Cartesian-grid spectral sampling, thereby removing constraints imposed by orthogonality and polar coordinates. Moreover, the framework accommodates non-strictly-orthogonal filters (e.g., Gabor), significantly enhancing robustness and generality. The result is a geometrically transparent, computationally tractable, and engineering-practical paradigm for structural tensor theory—offering greater simplicity, clearer geometric interpretation, and broader applicability than prior approaches.

Technology Category

• Machine Learning: Matrix & Tensor Methods
• Computer Vision: Other Foundations of Computer Vision
• Search and Optimization: Non-convex Optimization

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Web Mining and Content Analysis: Bridging structured and unstructured data

Abstract

This note presents a theoretical discussion of two structure tensor constructions: one proposed by Bigun and Granlund 1987, and the other by Granlund and Knutsson 1995. At first glance, these approaches may appear quite different--the former is implemented by averaging outer products of gradient filter responses, while the latter constructs the tensor from weighted outer products of tune-in frequency vectors of quadrature filters. We argue that when both constructions are viewed through the common lens of Total Least Squares (TLS) line fitting to the power spectrum, they can be reconciled to a large extent, and additional benefits emerge. From this perspective, the correction term introduced in Granlund and Knutsson 1995 becomes unnecessary. Omitting it ensures that the resulting tensor remains positive semi-definite, thereby simplifying the interpretation of its eigenvalues. Furthermore, this interpretation allows fitting more than a single 0rientation to the input by reinterpreting quadrature filter responses without relying on a structure tensor. It also removes the constraint that responses must originate strictly from quadrature filters, allowing the use of alternative filter types and non-angular tessellations. These alternatives include Gabor filters--which, although not strictly quadrature, are still suitable for structure tensor construction--even when they tessellate the spectrum in a Cartesian fashion, provided they are sufficiently concentrated.

Problem

• Reconciling two structure tensor constructions through TLS line fitting
• Eliminating correction terms to ensure positive semi-definite tensors
• Enabling alternative filter types beyond strict quadrature filters

Innovation

• Reconciles two tensor methods via TLS framework
• Eliminates correction term for positive semi-definite tensors
• Enables non-quadrature filters and Cartesian tessellations