Multivariate Fields of Experts

2025-08-08

AI Summary

This work addresses inverse problems in computational imaging—including image denoising, deblurring, compressed sensing MRI, and CT reconstruction—by proposing the Multivariate Field Expert Model (MV-EM). MV-EM explicitly encodes high-order image priors via a Moreau envelope-based multivariate potential function constructed from the ℓ∞-norm. Unlike conventional univariate field models, MV-EM is the first to extend the field expert framework to multivariate potential spaces, thereby significantly enhancing modeling capacity and generalization while preserving physical interpretability. The method operates within a classical regularization framework—requiring no deep neural networks—and enables lightweight training and inference through closed-form or iterative optimization. Experiments demonstrate that MV-EM matches or exceeds state-of-the-art deep learning methods across multiple tasks, while reducing model parameters by over 90%, decreasing training data requirements by an order of magnitude, and accelerating inference by 2–5×.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multi-instance/Multi-view Learning
• Search and Optimization: Non-convex Optimization

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data

Abstract

We introduce the multivariate fields of experts, a new framework for the learning of image priors. Our model generalizes existing fields of experts methods by incorporating multivariate potential functions constructed via Moreau envelopes of the $ell_infty$-norm. We demonstrate the effectiveness of our proposal across a range of inverse problems that include image denoising, deblurring, compressed-sensing magnetic-resonance imaging, and computed tomography. The proposed approach outperforms comparable univariate models and achieves performance close to that of deep-learning-based regularizers while being significantly faster, requiring fewer parameters, and being trained on substantially fewer data. In addition, our model retains a relatively high level of interpretability due to its structured design.

Problem

• Learning image priors via multivariate fields of experts
• Improving performance in inverse imaging problems
• Balancing speed, interpretability, and data efficiency

Innovation

• Multivariate potential functions via Moreau envelopes
• Generalizes fields of experts with l∞-norm
• Efficient interpretable model outperforming univariate