Variational Augmentation for Enhancing Historical Document Image Binarization

TL;DR

This paper introduces a two-stage deep learning framework that uses variational inference to generate degraded training samples, significantly improving historical document image binarization performance on challenging datasets.

Contribution

The novel two-stage framework combines variational inference-based data augmentation with CNN binarization, addressing data scarcity in historical document processing.

Findings

Achieved competitive results on DIBCO datasets

Outperformed several existing binarization methods

Demonstrated robustness on severely degraded images

Abstract

Historical Document Image Binarization is a well-known segmentation problem in image processing. Despite ubiquity, traditional thresholding algorithms achieved limited success on severely degraded document images. With the advent of deep learning, several segmentation models were proposed that made significant progress in the field but were limited by the unavailability of large training datasets. To mitigate this problem, we have proposed a novel two-stage framework -- the first of which comprises a generator that generates degraded samples using variational inference and the second being a CNN-based binarization network that trains on the generated data. We evaluated our framework on a range of DIBCO datasets, where it achieved competitive results against previous state-of-the-art methods.