Citation

Patil G, Shivakumara P, Gornale SS, Pal U, Blumenstein M. Multimed. Tools Appl. 2023; 82(14): 20925-20949.

Copyright

(Copyright © 2023, Holtzbrinck Springer Nature Publishing Group)

DOI

10.1007/s11042-022-14242-8

PMID

unavailable

Abstract

Altering handwritten documents is receiving special attention from researchers because it is useful in several sensitive crime applications such as the identification of suicide notes, fraudulent certificates, fake answer scripts, bank and property documents etc. This paper aims at developing a robust method for detecting altered text in handwritten document images in noisy and blurry environments. This study considers ten classes of handwritten text affected by multiple forgery operations with noise and blur as follows: (i) Normal-original text, (ii) Copy-paste forgery, (iii) Insertion forgery, (iv) Copy-paste and insertion forgery, (v) Noisy text, (vi) Blurred text, (vii) Copy-paste forgery with noise, (viii) Copy-paste forgery with blur, (ix) Insertion forgery with noise and (x) Insertion with blur. For ten-class classification, the proposed work explores the combination of statistical, gradient and texture features with a Bayesian classifier. The proposed approach works based on the premise that altered content in noisy and blurry handwritten documents exhibits inconsistent patterns of pixel arrangements while the original text exhibits a regular pattern of pixel arrangements. Comprehensive experiments on our dataset of 10-class and three standard datasets, namely, a dataset of forged handwritten text, a dataset of altered receipt images, and a dataset of forged IMEI number images are conducted to show effectiveness and robustness of the proposed approach compared to the state-of-the-art methods. © 2022, The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.

Language: en

Keywords

Textures; Classification (of information); Altered text detection; Copy-paste forgery; Document forgery; Fake detection; Handwritten document; Handwritten texts; Histogram of oriented gradients; Local binary pattern; Local binary patterns; Naive Bayesian classifier; Naive Bayesian Classifier; Pixels; Statistical features; Text detection