Archival handwriting degradation recovery method using cross-band information analysis

By using a cross-band information analysis method, grayscale images of archives in multiple narrowband bands are acquired, a three-dimensional data cube is constructed, and the separation matrix is ​​optimized using a sliding delay operator and multi-resolution discrete wavelet transform. This solves the nonlinear spectral mixing problem in the separation of handwriting in double-sided printed archives and achieves clear restoration of the handwriting.

CN122415384APending Publication Date: 2026-07-17
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Filing Date
2026-04-23
Publication Date
2026-07-17

AI Technical Summary

Technical Problem

In existing technologies, the toner on the front side of double-sided printed documents is embedded in the paper fibers due to long-term compression and forms a nonlinear spectral mixture with the ballpoint pen ink on the back side. This causes traditional linear blind source separation methods to be unable to effectively separate the handwriting in the documents, and the separation results are highly sensitive to the initial values ​​and are prone to producing false contours.

Method used

A cross-band information analysis method is adopted. By acquiring grayscale images of archives in multiple narrowband bands, a three-dimensional data cube is constructed. Time-frequency transformation is performed using the sliding delay operator and multi-resolution discrete wavelet transform. Combined with iterative update rules, the separation matrix is ​​optimized to separate the front and back handwriting and paper background signals. The feature grayscale matrix is ​​then selected based on the feature absorption band for restoration.

Benefits of technology

It achieves handwriting separation without afterimages, restores the continuous and complete strokes of the handwriting in the image, with clear edges and no crosstalk, avoids manual screening, and the final output handwriting image is clear and without false outlines.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122415384A_ABST
    Figure CN122415384A_ABST
Patent Text Reader

Abstract

This invention discloses a method for restoring degraded archival handwriting using cross-band information analysis. The method involves acquiring grayscale images of the archive to be restored in multiple narrowband bands and stacking them into a three-dimensional data cube. The spectral curves of each spatial location are extracted, and a sliding delay operator is used for convolution to obtain a mixed observation sequence. A discrete wavelet transform is applied to the mixed observation sequence, and an initial separation matrix is ​​constructed at each frequency point. This separation matrix is ​​then iteratively optimized by minimizing the trace of the auxiliary variable. After applying the separation matrix to the mixed observation sequence, an inverse wavelet transform is used to reconstruct the cross-band data cubes representing the front and back of the handwriting, as well as the paper background. Feature absorption bands are selected based on the product of variance and edge change intensity. A two-dimensional grayscale matrix is ​​extracted through linear interpolation of adjacent bands, and the restored handwriting image is output. This invention stably separates double-sided overlapping handwriting under nonlinear mixing conditions and eliminates false contours.
Need to check novelty before this filing date? Find Prior Art