Abstract:Preprocessing for degraded document images can improve their binarization result. Sharpening and illumination compensation are effective methods in preprocessing. We find the degree of sharpening has different effects on the different stroke width. As the degree of sharpening increases, the thin strokes retained more information in the binarization results, while the thick strokes gradually appear to be broken. Aiming at this problem, we propose a local adaptive sharpening method. The stroke width estimation algorithm is utilized to estimate the stroke width in the local region. Local adaptive sharpening is performed to solve the problem of thick stroke fracture and retain more information of thin strokes. In addition, comparing with the weakly illuminated document images, the sharpening effects on strongly illuminated document images are more prominent, and the binarization result is better. Therefore, appropriate illumination compensation is used for weakly illuminated document images. We further propose a preprocessing method for degraded document image using local adaptive sharpening and illumination compensation. The experimental results show that our proposed method restores more detail information and keeps the thick stroke information in binarization result. Our method outperforms U-Net without preprocessing by 0.36% FM scores on DIBCO2016, 1.09% FM scores on DIBCO2017 and 1.42% FM scores on DIBCO2018. U-Net, Sauvola and OTSU combined with our LASIC outperform themselves by 1.42%, 0.29% and 5.41% FM scores on DIBCO2018. And our LASIC method outperforms other preprocessing methods by 0.1% to 1.05% FM scores on DIBCO2016-DIBCO2018.

DeepOtsu: Document enhancement and binarization using iterative deep learning

An Iterative Refinement Framework for Image Document Binarization with Bhattacharyya Similarity Measure

Deep Networks for Degraded Document Image Binarization through Pyramid Reconstruction

A multi-scale framework for adaptive binarization of degraded document images

Adaptive Binarization for Degraded Document Image via Contrast Enhancement

Document Image Binarization with Feedback for Improving Character Segmentation

A Fair Evaluation of Various Deep Learning-Based Document Image Binarization Approaches

Adaptive degraded document image binarization

Parallel Nonparametric Binarization for Degraded Document Images.

An enhanced binarization framework for degraded historical document images

Degraded document image preprocessing using local adaptive sharpening and illumination compensation

BiNet: Degraded-Manuscript Binarization in Diverse Document Textures and Layouts using Deep Encoder-Decoder Networks

Adaptive Binarization for Degraded Document Images

A Novel Degraded Document Binarization Model Through Vision Transformer Network

A Novel Binarization System for Degraded Document Images.

Improving Document Binarization Via Adversarial Noise-Texture Augmentation

Document Image Binarization with Fully Convolutional Neural Networks

Fast document image binarization based on an improved adaptive Otsu's method and destination word accumulation

Joint Demosaicing and Denoising with Double Deep Image Priors

Evaluating Deep Neural Networks for Image Document Enhancement

Development and quantitative assessment of deep learning-based image enhancement for optical coherence tomography