High dynamic range imaging method based on space-frequency interaction

By employing a high dynamic range imaging method based on spatial-frequency interaction, combined with multi-scale convolution, attention gating, and frequency domain correction, the problems of ghosting artifacts and detail degradation in dynamic and complex scenes are solved, achieving efficient and stable image reconstruction, which is suitable for photography, autonomous driving, and remote sensing.

CN122156023APending Publication Date: 2026-06-05NORTHWESTERN POLYTECHNICAL UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
NORTHWESTERN POLYTECHNICAL UNIV
Filing Date
2025-12-12
Publication Date
2026-06-05

AI Technical Summary

Technical Problem

Existing high dynamic range imaging methods are prone to ghosting artifacts and detail degradation when reconstructing images in dynamic and complex scenes. Furthermore, traditional methods rely on insufficient spatial domain modeling, and frequency domain features are not fully utilized.

Method used

A high dynamic range imaging method based on spatial-frequency interaction is adopted. Local texture and edge details are captured by multi-scale convolution and attention gating. Combined with frequency domain amplitude and phase correction, a cross-domain attention mask is generated. A lightweight multilayer perceptron is used for feature fusion and cue optimization to achieve deep collaborative modeling of the spatial and frequency domains.

Benefits of technology

It significantly improves image quality stability and detail fidelity, solves ghosting artifacts and spatial misalignment problems in dynamic scenes, and balances model efficiency and scalability, making it suitable for fields such as photography, autonomous driving and remote sensing.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122156023A_ABST
    Figure CN122156023A_ABST
Patent Text Reader

Abstract

The application particularly relates to a high dynamic range imaging method based on space-frequency interaction, which comprises the following steps: taking a low dynamic range image as input, generating a corresponding high dynamic range image through gamma correction, processing the high dynamic range image through a shared convolution layer after splicing, and generating initial alignment features by using an attention-based alignment method; respectively performing space domain and frequency domain feature extraction on the initial alignment features; adaptively balancing the contributions of the space domain and frequency domain features through a cross-domain feature fusion block to generate fusion features; extracting key structure clues from the fusion features by using a prompt optimization module, and directionally repairing degradation areas such as oversaturation and misplacement; and finally outputting a high-fidelity high dynamic range image. The application fully gives play to the complementary advantages of space-frequency dual domains, and combines an efficient attention mechanism and a learnable prompt template, so that the image structure consistency and detail integrity can still be guaranteed in an extreme light and large motion scene.
Need to check novelty before this filing date? Find Prior Art