ItemArticle

Convolutional bi-directional autoencoder assisted generative adversarial de-blurring framework for palm leaf character analysis

Title
Convolutional bi-directional autoencoder assisted generative adversarial de-blurring framework for palm leaf character analysis
Creator
SARITHADEVI, S.; RAJESH, R.
Description
Palm leaf manuscripts have historically educated people on a variety of topics, including astronomy, mathematics, astrology and medicine. These manuscripts are constructed from dried palm leaves, which contain a wealth of information that remains largely untapped due to the challenges of digitalization and transcription. Recognizing the characters found in palm leaf manuscripts is a complex problem because blurred images of these manuscripts often conceal critical information. The present study proposes an automated de-blurring model to effectively identify exact Malayalam characters in palm leaf manuscripts. The input images are gathered from a real-time dataset to address this challenge. The Weighted Guided Image Filtering (WG_IF) method is employed to extract the detail layer, which not only reveals important information about the characters but also eliminates unwanted noise. The detailed layer is then input into the proposed Convolutional Bi-directional AutoEncoder assisted Generative Adversarial De-blurring (CBiAE_GADeblur) framework. This framework comprises two main blocks: the generator and the discriminator. The generator block uses the Convolutional Bi-directional Long Short Term Memory with AutoEncoder (CBLSTM_AE) method to produce de-blurred images. The discriminator block classifies these images as real or fake, enhancing the prediction accuracy of the palm characters. The proposed method demonstrates a superior accuracy rate of 98.25% in Prasavachikilsa and Vishavydyam datasets and exhibits lower time complexity. The motivation behind this research is to overcome the significant barriers posed by blurred palm leaf images, thereby unlocking and preserving the invaluable knowledge contained within these historical documents. Indian Academy of Sciences 2025.
Source
Sadhana - Academy Proceedings in Engineering Sciences;Volume;50;Issue;4;Article No.;269;
Date
01-01-2025
Publisher
Springer
Subject
AutoEncoder; Bi-LSTM; CNN; discriminator; GANet; Guided image filter; RNN
Coverage
SARITHADEVI S., Department of Computer Science, CHRIST (Deemed to be University), Karnataka, Bengaluru, 560029, India, Department of Computer Science, Naipunnya Institute of Management and Information Technology, Kerala, Thrissur, 680308, India; RAJESH R., Department of Statistics and Data Science, CHRIST (Deemed to be University), Karnataka, Bengaluru, 560029, India
Rights
Restricted Access; Hardcopy may be available in the library
Relation
ISSN: 2562499; CODEN: SAPSE
Format
online
Language
English
Type
Article

Citation

SARITHADEVI, S.; RAJESH, R., “Convolutional bi-directional autoencoder assisted generative adversarial de-blurring framework for palm leaf character analysis,” CHRIST (Deemed To Be University) Institutional Repository, accessed June 19, 2026, https://archives.christuniversity.in/items/show/21994.

Output Formats