Segmentation of handwritten interference marks using multiple directional stroke planes and reformalized...

Area of Science:

• Computer Science
• Image Processing
• Document Analysis

Background:

• Printed documents often contain interference marks, such as strokes and smudges, which hinder accurate text extraction.
• Conventional morphological operations can suffer from the 'flooding water' effect, leading to errors in word segmentation.

Purpose of the Study:

• To present a new algorithm for robust word extraction from degraded printed documents.
• To address the limitations of existing methods in handling interference marks and similar text obstructions.

Main Methods:

• The proposed algorithm utilizes morphological operations informed by multiple direction projection planes.
• Skeleton images are employed to refine the segmentation process and prevent the 'flooding water' artifact.
• The approach focuses on accurately isolating individual words despite overlapping or intersecting strokes.

Main Results:

• The algorithm demonstrated successful word extraction in the presence of significant interference.
• Qualitative and quantitative assessments confirmed the method's ability to overcome common segmentation challenges.
• Test results indicate a high degree of feasibility for the proposed word extraction technique.

Conclusions:

• The developed algorithm offers a viable solution for extracting text from challenging document images.
• The novel application of projection planes and skeletonization effectively mitigates interference-related segmentation errors.
• This approach contributes to improved optical character recognition (OCR) preprocessing for degraded historical or damaged documents.