CNN Based Binarization of MultiSpectral Document Images

Hollaus, Fabian; Brenner, Simon; Sablatnig, Robert

doi:10.1109/icdar.2019.00091

Cited by 8 publications

(4 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The regions are sampled from 48 different manuscripts and 8 language families. Slavonic (19), Latin (13) and Greek (12) texts make up the majority of the samples; additionally, two Ottoman texts and one each in Armenian, Georgian, German and Gothic are contained in the dataset. Depending on line height and layout, the regions contain 1-17 lines of text.…”

Section: A Test Imagesmentioning

confidence: 99%

“…The manuscript regions represented in the SALAMI dataset are drawn from a set of approximately 4600 pages of 67 historical manuscripts, of which the Computer Vision Lab (TU Wien) has acquired multispectral images in the course of consecutive research projects between 2007 and 2019 1 . The multispectral images were acquired using different imaging devices and protocols [13], [25], [26], with 6-12 wavebands per image.…”

Section: A Test Imagesmentioning

confidence: 99%

“…This approach addresses the property of readability more directly than the approaches mentioned before and is a reasonable choice when the purpose of text restoration lies in the subsequent processing with a HCR system, rather than in the preparation for a human observer. For restoration approaches producing binary images, evaluation is more straight-forward: for this purpose, multispectral image datasets with ground truth annotations have been published [12], [13]. While the above-named approaches to quantitative evaluation are promising for their respective use cases, their correlation to legibility by human observers is yet to be shown.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Subjective Assessments of Legibility in Ancient Manuscript Images - The SALAMI Dataset

Brenner

Sablatnig

2021

Pattern Recognition. ICPR International Workshops and Challenges

Self Cite

View full text Add to dashboard Cite

The research field concerned with the digital restoration of degraded written heritage lacks a quantitative metric for evaluating its results, which prevents the comparison of relevant methods on large datasets. Thus, we introduce a novel dataset of Subjective Assessments of Legibility in Ancient Manuscript Images (SALAMI) to serve as a ground truth for the development of quantitative evaluation metrics in the field of digital text restoration. This dataset consists of 250 images of 50 manuscript regions with corresponding spatial maps of mean legibility and uncertainty, which are based on a study conducted with 20 experts of philology and paleography. As this study is the first of its kind, the validity and reliability of its design and the results obtained are motivated statistically: we report a high intra-and inter-rater agreement and show that the bulk of variation in the scores is introduced by the images regions observed and not by controlled or uncontrolled properties of participants and test environments, thus concluding that the legibility scores measured are valid attributes of the underlying images.

show abstract

Section: A Test Imagesmentioning

confidence: 99%

Section: A Test Imagesmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Subjective Assessments of Legibility in Ancient Manuscript Images - The SALAMI Dataset

Brenner

Sablatnig

2021

Pattern Recognition. ICPR International Workshops and Challenges

Self Cite

View full text Add to dashboard Cite

show abstract

“…Thus, we consider the MSIO (MultiSpectral Document Image BinarizatIOn) method proposed by Diem et al [21,26], and the method proposed by Hollaus et al [21]. We also consider the results of a recent CNN-based approach proposed by Hollaus et al [27]. As we can see in Table 2, a significant improvement in terms of all metrics is obtained by the proposed KONMF algorithm over all other benchmark methods, including the two winners of the MS-TEx 2015 contest, which demonstrates the efficiency of the proposed KONMF approach for the decomposition task.…”

Section: Resultsmentioning

confidence: 99%

Kernel Orthogonal Nonnegative Matrix Factorization: Application to Multispectral Document Image Decomposition

Rahiche

Cheriet

2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

As a nonlinear extension of the standard nonnegative matrix factorization (NMF), kernel-based variants have demonstrated to be more effective for discovering meaningful latent features from raw data. However, many existing kernel methods allow only obtaining the basis matrix in the projected feature space, which prevents its inverse mapping back to the original space as requested in many applications. In this work, we propose a new kernel orthogonal NMF method that does not suffer from the pre-image issue. We incorporate the orthogonality constraint as an optimization problem over the Stiefel manifold to improve the sparsity and the model's clustering properties. We solve the proposed model with an efficient optimization approach based on the alternating direction method of multipliers (ADMM) scheme and the projected gradients method. We validate our model on the task of blind decomposition of real-world Multispectral (MS) document images. Our experiments demonstrate the competitiveness of our proposed model in comparison to the stateof-the-art techniques.

show abstract

Approach for the Classification of Polliniferous Vegetation Using Multispectral Imaging and Neural Networks

Silva

Varela

Díaz-Martinez

et al. 2020

Advances in Intelligent Systems and Computing

View full text Add to dashboard Cite

Beekeeping has suffered a serious deterioration in the regions of the world. The possibility of nectar-polliniferous resources has decreased and, therefore, the feeding of bees, with the consequent decrease in production. There is, therefore, a need to improve marketing and diversification systems and to update production techniques and the management of the production process by beekeepers to obtain the quality of honey required by the market [1]. This work proposes the use of spectral information to identify the different pollen-producing plants using remote vision, image processing, and artificial neural networks.

show abstract

CNN Based Binarization of MultiSpectral Document Images

Cited by 8 publications

References 11 publications

Subjective Assessments of Legibility in Ancient Manuscript Images - The SALAMI Dataset

Subjective Assessments of Legibility in Ancient Manuscript Images - The SALAMI Dataset

Kernel Orthogonal Nonnegative Matrix Factorization: Application to Multispectral Document Image Decomposition

Approach for the Classification of Polliniferous Vegetation Using Multispectral Imaging and Neural Networks

Contact Info

Product

Resources

About