Paired Image to Image Translation for Strikethrough Removal From Handwritten Words

TL;DR

This paper explores neural network models for removing strikethrough annotations from handwritten words, improving accuracy and efficiency over existing methods by using paired image translation techniques.

Contribution

It introduces and compares four neural network architectures for strikethrough removal, demonstrating superior performance and fewer parameters than CycleGAN-based approaches.

Findings

Paired models outperform CycleGAN in accuracy.

Proposed models use less than a sixth of the parameters.

Experimental results on synthetic and real datasets validate effectiveness.

Abstract

Transcribing struck-through, handwritten words, for example for the purpose of genetic criticism, can pose a challenge to both humans and machines, due to the obstructive properties of the superimposed strokes. This paper investigates the use of paired image to image translation approaches to remove strikethrough strokes from handwritten words. Four different neural network architectures are examined, ranging from a few simple convolutional layers to deeper ones, employing Dense blocks. Experimental results, obtained from one synthetic and one genuine paired strikethrough dataset, confirm that the proposed paired models outperform the CycleGAN-based state of the art, while using less than a sixth of the trainable parameters.