Improving Document Binarization Via Adversarial Noise-Texture Augmentation

Bhunia, Ankan Kumar; Bhunia, Ayan Kumar; Sain, Aneeshan; Roy, Partha Pratim

doi:10.1109/icip.2019.8803348

Cited by 28 publications

(27 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Also, after the introduction of generative models, image binarization systems with generative models have been suggested [23]. For enhancing the performance of image binarization, Bhunia et al [24] and Zhao et al [25] develop binarization systems using a conditional GAN (cGAN).…”

Section: A Image Binarization and Background Blurring Methodsmentioning

confidence: 99%

Tunable U-Net: Controlling Image-to-Image Outputs Using a Tunable Scalar Value

2021

View full text Add to dashboard Cite

Image-to-image conversion tasks are more accurate and sophisticated than ever thanks to advances in deep learning. However, since typical deep learning models are trained to perform only one task, multiple trained models are required to perform each task even if they are related to each other. For example, the popular image-to-image convolutional neural network, U-Net, is normally trained for a single task. Based on U-Net, this study proposes a model that outputs variable results using only one trained model. The proposed method produces a continuously changing output by setting an external parameter. We confirm the robustness of our proposed model by evaluating it on binarization and background blurring. According to these evaluations, we confirmed that the proposed model can generate well-predicted outputs by using un-trained tuning parameters as well as the outputs by using trained tuning parameters. Furthermore, the proposed model can generate extrapolated outputs outside the learning range.INDEX TERMS Image-to-image conversion, multiple tasks, U-Net, image binarization, background blur

show abstract

Section: A Image Binarization and Background Blurring Methodsmentioning

confidence: 99%

Tunable U-Net: Controlling Image-to-Image Outputs Using a Tunable Scalar Value

2021

View full text Add to dashboard Cite

show abstract

“…We can solve classification tasks, for example, identifying diseases in plants [10], and even group images based on its pixel contents for effective image retrieval from large databases, just as implemented in [11]. Image generation tasks like document binarization [12] are more advanced, and adversarial networks can be implemented. Super resolution falls into the category of image generation, as the output is also an image, but with a higher resolution.…”

Section: Related Workmentioning

confidence: 99%

Efficient Multi-scalable Network for Single Image Super Resolution

Alao¹,

Kim²,

Kim³

et al. 2021

J Multimed Inf Syst

View full text Add to dashboard Cite

Studies show that pre-up-sampling techniques induce a significant number of operations compared to post-upsampling SR frameworks. In this paper, we introduce an efficient multi-scalable convolutional neural network constituting post-up-sampling but interpolation-based upscale technique. Similar to BTSRN [9], the proposed network consists of convolutional layers applied on the low-resolution input image and its feature maps, an upsampling layer, and more convolutional layers applied on the up-scaled feature maps. Unlike BTSRN [9], in the upscale layer, the multiple-channeled feature map of the previous layer is up-scaled by bicubic interpolation inducing multi-scalability. Additionally, unlike other previous works of which the training is performed using only scale factors of 2, 3, and 4, the proposed network is

show abstract

“…Similarly, a GAN discriminator is used to improve a prediction model in [178]. Bhunia et al [14] proposed a unique approach to train a binarization network using unpaired training data (i.e., the grayscale and binary images do not correspond) and achieved an impressive 97.8% FM on DIBCO13. To reduce the amount of needed training data, Krantz and Westphal [68] proposed a clustering method to ensure only diverse data are labeled.…”

Section: Deep Neural Networkmentioning

confidence: 99%

“…There is also limited labeled training data for deep models. Though some initial explorations around this issue have been made [14,55,68,165], there is still a large gap to address.…”

Section: Technical Challengesmentioning

confidence: 99%

“…Future work in the area might explore how to integrate some of the best practices from image processing techniques with deep learning models or vice versa, for example incorporating preprocessing and thresholding into the learning model or using deep models to replace many of the parameters or fixed constants in traditional methods. A large dataset suitable for data-hungry deep models with high-quality, low-bias pixel annotations also remains elusive, though GAN-based methods seem to help alleviate this data need [14].…”

Section: Directions For Future Workmentioning

confidence: 99%

See 1 more Smart Citation

Historical Document Image Binarization: A Review

Tensmeyer

Martinez

2020

SN COMPUT. SCI.

View full text Add to dashboard Cite

This review provides a comprehensive view of the field of historical document image binarization with a focus on the contributions made in the last decade. After the introduction of a standard benchmark dataset with the 2009 Document Image Binarization Contest, research in the field accelerated. Besides the standard methods for image thresholding, preprocessing, and post-processing, we review the literature on methods such as statistical models, pixel classification with learning algorithms, and parameter tuning. In addition to reviewing binarization algorithms, we discuss the available public datasets and evaluation metrics, including those that require pixel-level ground truth and those that do not. We conclude with recommendations for future work.

show abstract

Improving Document Binarization Via Adversarial Noise-Texture Augmentation

Cited by 28 publications

References 18 publications

Tunable U-Net: Controlling Image-to-Image Outputs Using a Tunable Scalar Value

Tunable U-Net: Controlling Image-to-Image Outputs Using a Tunable Scalar Value

Efficient Multi-scalable Network for Single Image Super Resolution

Historical Document Image Binarization: A Review

Contact Info

Product

Resources

About