Image-to-Video Synthesis Using Generative Adversarial Networks and the Poisson Equation
Technical Annals · 26.01.2026 · 2 (3)
DOI: 10.12681/ta.45925 ↗Kurz gefasst
Fachartikel in der Zeitschrift Technical Annals, online erschienen am 26.01.2026. Verfasst von Christophoros Nikou und 2 weiteren Autor:innen. Die Arbeit nennt 23 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-SA 4.0).
Abstract (Originalsprache)
Deep generative models have revolutionized video synthesis by enabling the animation of staticsource images through the extraction of motion patterns from driving sequences. Despite theseadvancements, a persistent challenge in the field remains the mitigation of structural degradation, where generated outputs frequently exhibit high-frequency frame noise and boundary artifacts that undermine visual quality. Adopting the First Order Motion Model (FOMM) as a baseline, we propose a post-processing pipeline that integrates the Poisson equation-based seamless cloning technique, guided by estimated occlusion maps and Gaussian filtering, to suppress synthesis noise and smooth edge discontinuities. To maintain temporal consistency and prevent the blending of background elements into the moving foreground, we further in troduce a custom thresholding mask, as the main part in applying the Poisson-based technique. Our experimental results on the Taichi and Fashion datasets demonstrate that this approach preserves structural integrity more effectively than standard filtering, yielding substantial improvements in Structural Similarity (SSIM) metrics.
Zitieren
APA
Nikou, C., Vrigkas, M., & Stergiou, V. (2026). Image-to-Video Synthesis Using Generative Adversarial Networks and the Poisson Equation. Technical Annals, 2(3). https://doi.org/10.12681/ta.45925
Vancouver
Nikou C, Vrigkas M, Stergiou V. Image-to-Video Synthesis Using Generative Adversarial Networks and the Poisson Equation. Technical Annals. 2026;2(3). doi:10.12681/ta.45925
BibTeX
@article{Nikou2026,
author = {Nikou, Christophoros and Vrigkas, Michalis and Stergiou, Vasileios},
title = {Image-to-Video Synthesis Using Generative Adversarial Networks and the Poisson Equation},
journal = {Technical Annals},
year = {2026},
volume = {2},
number = {3},
doi = {10.12681/ta.45925}
}
EUVIMED