• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권5호
• /
• pp.1778-1797
• /
• 2021

Document image binarization is an important pre-processing step in document analysis and archiving. The state-of-the-art models for document image binarization are variants of encoder-decoder architectures, such as FCN (fully convolutional network) and U-Net. Despite their success, they still suffer from three limitations: (1) reduced feature map resolution due to consecutive strided pooling or convolutions, (2) multiple scales of target objects, and (3) reduced localization accuracy due to the built-in invariance of deep convolutional neural networks (DCNNs). To overcome these three challenges, we propose an improved semantic segmentation model, referred to as DP-LinkNet, which adopts the D-LinkNet architecture as its backbone, with the proposed hybrid dilated convolution (HDC) and spatial pyramid pooling (SPP) modules between the encoder and the decoder. Extensive experiments are conducted on recent document image binarization competition (DIBCO) and handwritten document image binarization competition (H-DIBCO) benchmark datasets. Results show that our proposed DP-LinkNet outperforms other state-of-the-art techniques by a large margin. Our implementation and the pre-trained models are available at https://github.com/beargolden/DP-LinkNet.

• International Journal of Knowledge Content Development & Technology
• /
• 제5권2호
• /
• pp.11-24
• /
• 2015

It is common for archive libraries to provide public access to historical and ancient document image collections. It is common for such document images to require specialized processing in order to remove background noise and become more legible. Document images may be contaminated with noise during transmission, scanning or conversion to digital form. We can categorize noises by identifying their features and can search for similar patterns in a document image to choose appropriate methods for their removal. In this paper, we propose a hybrid binarization approach for improving the quality of old documents using a combination of global and local thresholding. This article also reviews noises that might appear in scanned document images and discusses some noise removal methods.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2003년도 신호처리소사이어티 추계학술대회 논문집
• /
• pp.393-396
• /
• 2003

This paper proposes a method to speed up the document image binarization using a water flow model. The proposed method extracts the region of interest (ROI) around characters from a document image and restricts pouring water onto a 3-dimensional terrain surface of an image only within the ROI. The amount of water to be filled into a local valley is determined automatically depending on its depth and slope. Then, the proposed method accumulates weighted water not only on the locally lowest position but also on its neighbors. Finally, the depth of each pond is adaptively thresholded for robust character segmentation. Experimental results on real document images shows that the proposed method has attained good binarization performance as well as remarkably reduced processing time compared with that of the existing method based on a water flow model.

• 정보처리학회논문지B
• /
• 제13B권4호
• /
• pp.369-376
• /
• 2006

디지탈 변환과 기기간의 전송 영향으로 화질이 떨어진 디지탈 영상의 복구는 잡음 발생 및 그 역 과정의 모형화를 통해 이루어낼 수 있다. 스캐너로 읽혀진 서류 영상이나 위성 사진에서 잡음 및 반점을 제거하는 과정이 좋은 예이다. 그러나 잡음 발생의 비선형성은 그 역 과정의 이론적 이해를 어렵게한다. 본 논문에서는 충격 잡음에의해 화질이 떨어진 이진 서류 영상의 복구 방법들을 심층 분석하는 것에 촛점을 맞추었다. 본 연구 결과에 의하면 이진 서류 영상의 잡음 제거 방식으로 '가중 중앙값' 여과기와 '리' 여과기가 다른 여과기에 비해 효과적임을 보여준다. 반면 '웨이브렛' 여과 방식은 타 방식보다 100여배의 시간이 소요되어 비효율적이다. 본 논문에서는 가중 중앙값 여과기에 쓰이는 가중치에 대한 연구 결과를 제시하였다.

• 정보처리학회논문지B
• /
• 제13B권6호
• /
• pp.591-600
• /
• 2006

본 논문에서는 문서 영상의 그림 영역에서 통계적 분석을 통한 단어 영상을 추출하는 방법을 제안한다. 제안 방법은 그림 영역의 구성 요소를 문자 성분과 그래픽 성분으로 분류하기 위하여 연결요소에 대한여 통계적 분석 방법인 상자그림 분석을 적용하고, 분류된 문자 성분들에 대하여 지역적 밀집도를 분석하여 문자 영역을 추출한다. 추출된 문자 영역에서 투영 히스토그램 분석을 통하여 문자열을 추출하고, 문자열을 단어단위 영상으로 분리하기 위하여 투영 히스토그램 분석과 갭 군집화, 특수 기호 검출 등을 수행한다. 제안 방법은 임계값의 사용 대신에 그림 영역의 구성 요소들에 대하여 통계적 분석을 수행하기 때문에 그림의 형태 변화에 민감하지 않으며, 지역적 밀집도 분석으로 보다 정확한 문자 영역을 추출하였다. 또한 제안 방법의 응용 분야인 주제어 검색을 위한 오프라인의 전처리에 해당하는 문서 영상의 단어단위 영상 추출에 적용하여 제안 방법에 대한 연구의 필요성을 제시하였다.

• International Journal of Contents
• /
• 제11권2호
• /
• pp.1-8
• /
• 2015

In this paper, we describe a novel method for document layout analysis that is based on a Fuzzy Energy Matrix (FEM). A FEM is a two-dimensional matrix that contains the likelihood of text and non-text and is generated through the use of Fuzzy theory. The key idea is to define an Energy map for the document to categorize text and non-text. The proposed mechanism is designed for execution with a low-resolution document image, and hence our method has a fast processing speed. The proposed method has been tested on public ICDAR 2009 datasets to conduct a comparison against other state-of-the-art methods, and it was also tested with Korean documents. The results of the experiment indicate that this scheme achieves superior segmentation accuracy, in terms of both precision and recall, and also requires less time for computation than other state-of-the-art document image analysis methods.

• 한국멀티미디어학회논문지
• /
• 제13권12호
• /
• pp.1786-1797
• /
• 2010

Automatic understanding of contents in document image is a very hard problem due to involvement with mathematically challenging problems originated mainly from the over-determined system induced by document segmentation process. In both academic and industrial areas, there have been incessant and various efforts to improve core parts of content retrieval technologies by the means of separating out segmentation related issues using semi-structured document, e.g., invoice,. In this paper we proposed classification models for text lines on invoice document in which text lines were clustered into the five categories in accordance with their contents: purchase order header, invoice header, summary header, surcharge header, purchase items. Our investigation was concentrated on the performance of machine learning based models in aspect of linear-discriminant-analysis (LDA) and non-LDA (logic based). In the group of LDA, na$\"{\i}$ve baysian, k-nearest neighbor, and SVM were used, in the group of non LDA, decision tree, random forest, and boost were used. We described the details of feature vector construction and the selection processes of the model and the parameter including training and validation. We also presented the experimental results of comparison on training/classification error levels for the models employed.

• 한국콘텐츠학회논문지
• /
• 제15권11호
• /
• pp.18-26
• /
• 2015

스캐너 대신 카메라를 이용하여 문서의 사본 영상을 촬영하면 촬영 각도에 따라 기하학적 왜곡이 발생하거나 그림자가 생길 수 있다. 본 논문에서는 카메라로 촬영한 문서 영상으로부터 왜곡을 보정하고 그림자 영향을 제거한 흑백 문서 영상 생성 알고리즘을 제안하였다. 카메라 렌즈의 방사 왜곡으로 인해 휘어진 테두리를 펴거나 촬영 각도에 따라 유입된 문서 외부 영역을 제거하기 위한 기하학적 보정을 위해 2차 미분 필터 기반의 문서 테두리 검출 방안을 마련하였다. 그리고 적응적 이진화 방법으로 그림자를 제거한 흑백 문서 영상을 생성하였다. 제안한 왜곡 보정 흑백 문서 영상 생성 알고리즘을 스마트 폰 카메라로 촬영한 문서 영상들을 대상으로 실험한 결과 우수한 처리 결과를 얻을 수 있었다.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송공학회 2013년도 추계학술대회
• /
• pp.237-238
• /
• 2013

The quality of document image captured from electronic display might be worse when it is compared with document image captured from paper. The problem appears because of Moir? noise. This problem can lead to achieve inaccurate intermediate result for further image processing. This paper proposes a method to remove Moir? noise of document images captured from electronic display. The proposed algorithm is separated in two parts. In the first step, it corrects the text area region (foreground) with small area of smoothing. Then, it corrects the background area with large area of smoothing.

• International Journal of Contents
• /
• 제13권4호
• /
• pp.70-79
• /
• 2017

Images are an important element in patents and many experts use images to analyze a patent or to check differences between patents. However, there is little research on image analysis for patents partly because image processing is an advanced technology and typically patent images consist of visual parts as well as of text and numbers. This study suggests two methods for using image processing; the Scale Invariant Feature Transform(SIFT) algorithm and Optical Character Recognition(OCR). The first method which works with SIFT uses image feature points. Through feature matching, it can be applied to calculate the similarity between documents containing these images. And in the second method, OCR is used to extract text from the images. By using numbers which are extracted from an image, it is possible to extract the corresponding related text within the text passages. Subsequently, document similarity can be calculated based on the extracted text. Through comparing the suggested methods and an existing method based only on text for calculating the similarity, the feasibility is achieved. Additionally, the correlation between both the similarity measures is low which shows that they capture different aspects of the patent content.