Vehicle-mounted image processing method and device, terminal equipment and storage medium

By using optical flow field-guided multi-scale feature alignment and spatiotemporal attention mechanism for information fusion, the problem of accuracy and real-time performance of image restoration in dynamic scenes by vehicle cameras is solved, achieving efficient image restoration results.

CN122156008APending Publication Date: 2026-06-05SHENZHEN STREAMING VIDEO TECH
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
SHENZHEN STREAMING VIDEO TECH
Filing Date
2026-01-20
Publication Date
2026-06-05

AI Technical Summary

Technical Problem

Vehicle-mounted cameras are susceptible to image degradation in dynamic scenes due to harsh environmental conditions, making it difficult to meet the requirements of real-time performance and accuracy. Existing technologies are unable to effectively repair this degradation.

Method used

By acquiring the video frame to be repaired and its adjacent video frames, an optical flow estimation network is used to calculate the optical flow field, perform multi-scale feature extraction and alignment, and combine it with a repair model based on spatiotemporal attention mechanism for information fusion and reconstruction, thereby improving the accuracy and real-time performance of image repair.

Benefits of technology

It significantly improves the accuracy of image restoration in dynamic in-vehicle scenarios, avoids the problems of target blurring and misalignment after restoration, meets the real-time requirements of in-vehicle systems, and provides reliable environmental perception support for intelligent driving.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122156008A_ABST
    Figure CN122156008A_ABST
Patent Text Reader

Abstract

The application is suitable for the technical field of intelligent terminals, and provides a vehicle-mounted image processing method and device, a terminal device and a storage medium. The method comprises the following steps: acquiring a to-be-repaired video frame and adjacent video frames thereof collected by a vehicle-mounted camera; calculating an optical flow field based on the to-be-repaired video frame and the adjacent video frames and an optical flow estimation network; performing multi-scale feature extraction on the to-be-repaired video frame and the adjacent video frames to obtain original multi-scale features of the to-be-repaired video frame and multi-scale features of the adjacent video frames; aligning the multi-scale features of the adjacent video frames to a coordinate system of the to-be-repaired video frame based on the optical flow field; and inputting the original multi-scale features of the to-be-repaired video frame and the aligned multi-scale features of the adjacent video frames into a repair model based on a space-time attention mechanism to perform information fusion and reconstruction, so as to obtain a repaired video frame. The application can improve the accuracy of image repair in a vehicle-mounted dynamic scene and meet the real-time requirement of a vehicle-mounted system.
Need to check novelty before this filing date? Find Prior Art