一种基于多模态采集的高帧率超分辨率的提升方法
By combining multimodal data acquisition and feature fusion of event cameras and traditional cameras, the problems of data volume and energy consumption in high frame rate and high resolution image acquisition in existing technologies are solved, and low-cost, high-quality video generation is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- GUANGDONG BOHUA UHD INNOVATION CENT CO LTD
- Filing Date
- 2022-09-07
- Publication Date
- 2026-07-17
AI Technical Summary
Existing super-resolution enhancement techniques mainly process single images or single modalities, ignoring the temporal characteristics of video content and failing to effectively utilize spatiotemporal information. This results in large data volumes and high energy consumption when acquiring high frame rate and high resolution images, and existing methods fail to fully leverage the advantages of event cameras.
Data is acquired simultaneously using event cameras and traditional cameras. Event data is reconstructed through a sliding window, and feature fusion is performed by combining a feature extraction network and a temporal attention model. High-resolution, high-frame-rate video is generated using an adversarial generative network, including optical flow features of event data for frame interpolation reconstruction.
It achieves high frame rate and high resolution video acquisition with low cost and low data volume, making full use of the spatiotemporal advantages of event cameras, and is suitable for high-quality video acquisition in complex scenarios.
Smart Images
Figure CN115393194B_ABST