Sparse Representation Based Inpainting for the Restoration of Document Images Affected by Bleed-Through

Hanif, Muhammad Imran; Tonazzini, Anna; Savino, Pasquale; Salerno, Emanuele

doi:10.3390/proceedings2020093

Cited by 10 publications

(3 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This is a well-studied problem with a long line of work. To solve this problem various methods were suggested that include thresholding, partial differential equations (PDE) applica-tion [18,27], sparse representation [15], deep neural networks [16] and optical density models [38], Markov Random Fields (MRF) [37], etc. Sometimes a byproduct of this process is a restored text image from the opposite side that can be considered from the perspective of a palimpsest as an under-text.…”

Section: Bleed-through Removal Modelsmentioning

confidence: 99%

Revealing a history: palimpsest text separation with generative networks

Starynska

Messinger

Kong

2021

IJDAR

View full text Add to dashboard Cite

A palimpsest is a historical manuscript in which the original text (termed under-text) was erased and overwritten with another script in order to recycle the parchment. One of the main challenges in studying palimpsests is to reveal the under-text. Due to the development of multi-spectral imaging, the original text can sometimes be recovered through material differences of inks and parchment (Easton et al., in: 2011 19th European signal processing conference, IEEE, 2011). However, generally, the revealed text can be observed only partially due to the overlap with newer text and degradation of the material. In this work, we propose revealing the under-text completely using deep generative networks, by leveraging prior spatial information of the under-text script. To optimize the under-text, we mimic the process of palimpsest creation. This is done by generating the under-text from a separately trained generative network to match it to the palimpsest image after mixing it with foreground text. The mixing process is represented by a separate neural network, that is optimized with the under-text image to match the original palimpsest. We also add an additional background generative network to compensate for the unevenness of the background. We propose a novel way of training the background generative network, that does not require isolated background samples and can use any region with layers of text. This paper illustrates the first known attempt to solve palimpsest text layer separation with deep generative networks. We evaluate our method performance on artificial and real palimpsest manuscripts by measuring character recognition and pixel-wise accuracy of the reconstructed under-text.

show abstract

Section: Bleed-through Removal Modelsmentioning

confidence: 99%

Revealing a history: palimpsest text separation with generative networks

Starynska

Messinger

Kong

2021

IJDAR

View full text Add to dashboard Cite

show abstract

“…Current research on decomposing images with overcomplete dictionaries (i.e. more atoms than the dimension of image) has applications that benefit from sparsity of representation, such as image denoising [27,28], single image super-resolution reconstruction [29,30], image inpainting [31,32], and image compression [33]. A review of sparse representation in dictionary learning and image fusion (by using multiple sensors) was carried out by Zhang et al [34].…”

Section: Introductionmentioning

confidence: 99%

Basis-updating for data compression of displacement maps from dynamic DIC measurements

Chang

Wang

Siebert

et al. 2019

Mechanical Systems and Signal Processing

View full text Add to dashboard Cite

The extraction of useful information and removal of redundant noise from data has become a major research topic in recent years. Data compression is necessary for all kinds of analysis, and the demand for efficient compression techniques has gained much attention. Digital image correlation is a camera-based measuring system, which has been widely applied in strain analysis because of the convenience of measuring displacement fields by simply selecting a region of interest. Currently, there is interest in applying such methods to engineering structures in dynamics. However, one of the major issues related to the integration of camera-based systems with dynamic measurement is the generation of huge amounts of data, typically extending to many thousands of data points, because of the requirements of high sampling rate, spatial resolution, and long duration of recording. In this paper a new algorithm is presented that addresses the need for efficiency in full-field data processing. By making use of the data itself and combining the concept of sparse representation with Gram-Schmidt orthogonalisation, the number of basis function used to represent the data can be reduced and a concise decomposition established. In both simulated and experimental cases, the compression ratios for data size and number of signals used in operational modal analysis are substantially diminished, thereby demonstrating the effectiveness of the proposed algorithm. A reduced number of new basis functions is determined for the representation of data under the condition that the reconstructed displacement map reproduces the raw measured data to within a chosen threshold on the coefficient of correlation.

show abstract

“…[1]These factors may effect on scientific and important cultural material. So to maintain these physical and ancient culture materials some of algorithms and ways are used to improve the quality and remove the clutter by using lots of filters on those materials [2].…”

Section: …………………………………………………………………………………………………… Introduction:-mentioning

confidence: 99%