Exportar Publicação

Maqsood, R., Nunes, P., Conti, C. & Soares, L. D. (2025). WaveE2VID: Frequency-aware event-based video reconstruction. 2025 IEEE International Conference on Image Processing (ICIP), Proceedings (pp. 570?575). IEEE. https://doi.org/10.1109/ICIP55913.2025.11084548

R. Maqsood et al.,  "WaveE2VID: Frequency-aware event-based video reconstruction", in 2025 IEEE Int. Conf. on Image Processing (ICIP), Proc., Anchorage, AK, USA, IEEE, 2025, pp. 570-575

@inproceedings{maqsood2025_1793210369934,
	author = "Maqsood, R. and Nunes, P. and Conti, C. and Soares, L. D.",
	title = "WaveE2VID: Frequency-aware event-based video reconstruction",
	booktitle = "2025 IEEE International Conference on Image Processing (ICIP), Proceedings",
	year = "2025",
	editor = "",
	volume = "",
	number = "",
	series = "",
	doi = "10.1109/ICIP55913.2025.11084548",
	pages = "570-575",
	publisher = "IEEE",
	address = "Anchorage, AK, USA",
	organization = "",
	url = "https://ieeexplore.ieee.org/document/11084548"
}

TY  - CPAPER
TI  - WaveE2VID: Frequency-aware event-based video reconstruction
T2  - 2025 IEEE International Conference on Image Processing (ICIP), Proceedings
AU  - Maqsood, R.
AU  - Nunes, P.
AU  - Conti, C.
AU  - Soares, L. D.
PY  - 2025
SP  - 570-575
SN  - 1522-4880
DO  - 10.1109/ICIP55913.2025.11084548
CY  - Anchorage, AK, USA
UR  - https://ieeexplore.ieee.org/document/11084548
AB  - Event cameras, which detect local brightness changes instead of capturing full-frame images, offer high temporal resolution and low latency. Although existing convolutional neural networks (CNNs) and transformer-based methods for event-based video reconstruction have achieved impressive results, they suffer from high computational costs due to their linear operations. These methods often require 10M-30M parameters and inference times of 30-110 ms per forward pass at a resolution of 640 × 480 on modern GPUs. Furthermore, to reduce computational costs, these methods apply CNN-based downsampling, which leads to the loss of fine details. To address these challenges, we propose an efficient hybrid model, WaveE2VID, which combines the frequency-domain analysis of the wavelet transform with the spatio-temporal context modeling of a deep convolutional recurrent network. Our model achieves 50% faster inference speed and lower GPU memory usage than CNN and transformer-based methods, maintaining reconstruction performance on par with state-of-the-art approaches across benchmark datasets.
ER  -