Method for estimating point spread function based on smooth closed contour binary pattern target and optical flow guide

By using a method based on a smooth closed contour binary pattern target and optical flow guidance, combined with an optical flow network and a neural network parameterized model, the shortcomings of existing PSF acquisition methods in terms of accuracy, efficiency and completeness are solved, and high-precision and robust PSF estimation is achieved, which is applicable to a variety of lenses and camera systems.

CN122244174APending Publication Date: 2026-06-19ZHEJIANG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
ZHEJIANG UNIV
Filing Date
2026-03-16
Publication Date
2026-06-19

AI Technical Summary

Technical Problem

Existing PSF acquisition methods struggle to achieve a balance between accuracy, efficiency, and completeness. Traditional algorithms struggle to achieve sub-pixel level accurate registration in cases of severe blurring, and traditional methods suffer from limitations in angle sampling and difficulties in geometric alignment.

Method used

A method based on a smooth closed contour binary pattern target and optical flow guidance is adopted. By combining optical flow network and neural network parameterization model, high-precision PSF estimation is achieved through target design, geometric alignment and optimization framework.

Benefits of technology

It achieves complete representation of spatial variations across the entire field of view, omnidirectional anisotropic sampling, and sub-pixel-level precise alignment, ensuring high accuracy and robustness of PSF estimation, and is suitable for a variety of lenses and camera systems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122244174A_ABST
    Figure CN122244174A_ABST
Patent Text Reader

Abstract

This invention discloses a point spread function (PSF) estimation method based on a smooth, closed-contour binary pattern target and optical flow guidance. The method includes: capturing images of a calibration target containing a smooth, closed-contour binary pattern using a target camera; generating a binary proxy image from the resulting blurred observation image; inputting both images into an optical flow estimation network; using the obtained optical flow field to perform differentiable distortion on the binary proxy image to generate a potentially sharp image; generating a PSF estimate using a point spread function neural network model; convolving the potentially sharp image and the PSF estimate, followed by Bayer sampling and demosaicing interpolation; calculating the loss with the demosaiced blurred observation image; jointly updating parameters through gradient backpropagation; and outputting the final PSF estimate after convergence. This invention effectively solves the problem of high-precision PSF estimation under complex aberrations by utilizing the continuous change in normal direction provided by the smooth contour and the flexible edge localization achieved by optical flow technology.
Need to check novelty before this filing date? Find Prior Art