Date: Thu. June 18, 2026 (15:30H)
Event: DeustoCCM Seminar
General Colloquium / Research-Oriented
Title: Optimal Image Transport over Sparse Dictionaries
Speaker: Junqing Huang
Affiliation: DeustoCCM, University of Deusto
Abstract. We propose a novel optimal image transport framework over sparse dictionaries for image-to-image translation tasks by integrating the complementary strengths of Sparse Representation (SR) and Optimal Transport (OT). Specifically, we design an optimization formulation in which individual image features (e.g., color and style, etc.) are encoded via sparse representation, while an optimal transport plan is inferred over the learned dictionaries in alignment with the encoding process. Due to the compactness of sparse coding and the reduced optimal transport over sparse dictionaries, the proposed optimization model is computationally tractable and empirically efficient, thereby providing a simple yet effective paradigm for simultaneous image representation and transformation. We demonstrate its effectiveness and versatility across two representative tasks: color transform and photo-realistic style transfer, and further show its potential for image super-resolution and video-based extensions. We also conduct a series of visual comparisons and quantitative evaluations to demonstrate visually plausible and photo-realistic transfer effects, showcasing competitive performance compared with state-of-the-art methods, including deep learning models.
WHERE
University of Deusto, DeustoTech.
Room: Logistar
Unibertsitate Etorb., 24, 48007 Bilbao, Bizkaia