Damage Detection in Handwritten Documents Using Intensity Histogram Clustering

View at publisher Summarise with VIRAT
Authors: B.L.P. Sankalpana, L. Ranathunga
Cited by: 0
References: 9

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

In Sri Lanka, Sinhala has recognized as the most ancient and widely used language. This research aims to identify damaged areas in degraded Sinhala handwritten documents. Several studies have been conducted in this field, with most focusing on detecting character damage using the gradient values of image pixels. Clustering based on gradients is effective for identifying edges and stroke consistency. However, damages such as tears, ink fading, stains, or smudges are often overlooked by this method. In this study, pixel intensity values are considered as the primary approach for damage detection. Additional techniques, including NLM filtering, histogram clustering, morphological processing, and stroke width estimation, are also discussed. The proposed method was evaluated on seventy-five aged document images that exhibited multiple types of damages within a single image, achieving an accuracy of 70.57 %, a precision of 74.19 %, a recall of 93.43 %, and an F1-score of 82.90 %.

Fields of Study & Subjects

Handwritten Text Recognition Techniques Image Processing and 3D Reconstruction Vehicle License Plate Recognition

Publication details

Journal / source
IEEE International Conference on Industrial and Information Systems (ICIIS)
Pages
442–447
Publication date
16 January 2026
Language
EN