A real-world image defogging method and system based on near-end scattering atmospheric reconstruction

By using a near-end scattering-based atmospheric reconstruction method, employing a unified atmospheric scattering physical model and staged reconstruction technology, and jointly estimating the atmospheric light field and transmittance, the problems of incomplete defogging and unnatural color reproduction in existing technologies are solved, achieving efficient and stable defogging effects in the real world.

CN122367799APending Publication Date: 2026-07-10TSINGHUA SHENZHEN INTERNATIONAL GRADUATE SCHOOL
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202610447597.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-04-07
Publication Date
2026-07-10

AI Technical Summary

Technical Problem

Existing single-image dehazing techniques suffer from incomplete dehazing, unnatural color reproduction, and insufficient stability across different scenes in the real world, making it difficult to effectively recover visual information in complex environments.

Method used

A near-end scattering-based atmospheric reconstruction method is adopted, which performs staged reconstruction through a unified atmospheric scattering physics model, jointly estimates the atmospheric light field, transmittance and clear scene radiation, uses neural networks for correction, and combines online non-uniform fog synthesis and selective self-distillation adaptive framework to achieve synergistic optimization of variables.

Benefits of technology

It improves the accuracy and reliability of single-image dehazing, ensures the consistency of dehazing results with real-world radiation information, enhances the stability and convergence of the algorithm, and can effectively handle dehazing tasks under complex lighting conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122367799A_ABST
    Figure CN122367799A_ABST
Patent Text Reader

Abstract

This invention relates to a real-world image dehazing method based on near-end scattering atmospheric reconstruction, comprising the following steps: receiving a single hazy image; performing staged reconstruction of the hazy image based on a unified atmospheric scattering physics model to jointly estimate the atmospheric light field, transmittance, and clear scene radiance that vary spatially with pixel location; wherein each stage of the reconstruction process includes: updating the atmospheric light field; updating the transmittance based on the updated atmospheric light field; updating the clear scene radiance based on the updated atmospheric light field and transmittance; and outputting the dehazing result based on the clear scene radiance. Through the above design, this application fundamentally achieves accuracy and reliability in dehazing single images in real-world complex scenes.
Need to check novelty before this filing date? Find Prior Art