The content introduces the concept of axial motion magnification to enhance the legibility of small motions in videos. It addresses the limitations of existing motion magnification methods and proposes a novel approach to magnify decomposed motions along user-specified axes. The content covers the introduction, related work, learning-based axial motion magnification, neural networks, training data generation, experiments, and an ablation study. It highlights the effectiveness of the proposed method in both axial and generic motion magnification scenarios.
A otro idioma
del contenido fuente
arxiv.org
Ideas clave extraídas de
by Kwon Byung-K... a las arxiv.org 03-27-2024
https://arxiv.org/pdf/2312.09551.pdfConsultas más profundas