Markov Random Field Based Text Identification from Annotated Machine Printed Documents

Peng, Xinhua; Setlur, Srirangaraj; Govindaraju, Venu; Sitaram, Ranganatha; Bhuvanagiri, Kiran Kumar

doi:10.1109/icdar.2009.237

Cited by 26 publications

(11 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Training on handwritten CCs and improved segmentation of touching characters and annotations (Fig. 8) would also likely improve precision/recall to match or exceed levels reported in [14] while competing with error rates reported in state-of-the-art approaches [18].…”

Section: Discussionmentioning

confidence: 79%

“…A SVM is used to isolate signatures from machine print in [12] and to identify sparse handwritten annotations occurring at arbitrary orientations in [13]. Kmeans clustering followed by MRF relabeling is used in [14] for segmentation of handwriting, machine print and noise with an overall recall of 96.33%. In [18] a novel approach to automatically discover features pushes error rates of handwriting and machine print to 13.8%.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Connected Component Level Discrimination of Handwritten and Machine-Printed Text Using Eigenfaces

Pinson¹,

Barrett

2011

2011 International Conference on Document Analysis and Recognition

View full text Add to dashboard Cite

We employ Eigenfaces to discriminate between handwritten and machine-printed text at the connected component (CC) level. Normalized images of machine print CCs are treated as points in a high-dimensional space. PCA yields a reduceddimensional character space. Representative machine print CCs are projected into character space and a local distance threshold for each representative is automatically determined. CCs are classified as machine print if they are within the local distance threshold of their closest machine print representative. Otherwise, they are classified as handwriting. Recursive character segmentation using min graph cut is used to address the problem of touching characters. Validation over a large NIST handwriting and machine print database demonstrates precision of 93.98% and 89.1% for machine print and handwriting respectively.

show abstract

Section: Discussionmentioning

confidence: 79%

Section: Introductionmentioning

confidence: 99%

Connected Component Level Discrimination of Handwritten and Machine-Printed Text Using Eigenfaces

Pinson¹,

Barrett

2011

2011 International Conference on Document Analysis and Recognition

View full text Add to dashboard Cite

show abstract

“…Patches that were generated as a single unit but contained both machine printed text and handwriting (including crosses, edit marks, and other annotations) within a single patch as shown in Fig.1 after using the dilation operation described in [18,14] on the HP Labs data set were divided into training and testing sets.…”

Section: Methodsmentioning

confidence: 99%

“…In [18], Zheng et al proposed a two step approach to identify three different types of patches (machine printed text, handwriting and noise) in mixed documents. Peng et al [14] used a Markov Random Field (MRF) based method to relabel three types of patches after initial classification. By projecting each word horizontally, Guo and Ma separated handwritten material from documents using a hidden Markov model [9].…”

Section: Introductionmentioning

confidence: 99%

Overlapped text segmentation using Markov random field and aggregation

Peng

Setlur

Govindaraju

et al. 2010

Proceedings of the 9th IAPR International Workshop on Document Analysis Systems

Self Cite

View full text Add to dashboard Cite

Separating machine printed text and handwriting from overlapping text is a challenging problem in the document analysis field and no reliable algorithms have been developed thus far. In this paper, we propose a novel approach for separating handwriting from binary image of overlapped text. Instead of using fixed size training patches, we describe an aggregation method which uses shape context features to extract training samples automatically. We use a Markov Random Field (MRF) to model the overlapped text. The neighbor system is inherited from a coarsening procedure and the prior and likelihood of the MRF is learned based on a distance metric. Experimental results show that the proposed method can achieve 87.97% recall for handwriting and 91.44% recall for machine printed text.

show abstract

“…Recent work in pixel-level labeling would seem to be applicable 7-9 because fragmentation of foreground markings into units smaller than connected components is essential to this task, while much academic work has stopped at the connected component level. 10,11 The classic segmentation/recognition dilemma of mainstream computer vision is operative here: given correctly segmented image layers, OCR and graphics recognition would be much more tractable, and conversely, if the text and graphical shapes were known in advance, it would be relatively easier to discover their poses and supporting pixel-level evidence in the image.…”

Section: Data Capture / Functional Role Labelingmentioning

confidence: 99%

Scientific challenges underlying production document processing

Saund¹

2011

SPIE Proceedings

View full text Add to dashboard Cite

The Field of Document Recognition is bipolar. On one end lies the excellent work of academic institutions engaging in original research on scientifically interesting topics. On the other end lies the document recognition industry which services needs for high-volume data capture for transaction and back-office applications. These realms seldom meet, yet the need is great to address technical hurdles for practical problems using modern approaches from the Document Recognition, Computer Vision, and Machine Learning disciplines. We reflect on three categories of problems we have encountered which are both scientifically challenging and of high practical value. These are Doctype Classification, Functional Role Labeling, and Document Sets. Doctype Classification asks, "What is the type of page I am looking at?" Functional Role Labeling asks, "What is the status of text and graphical elements in a model of document structure?" Document Sets asks, "How are pages and their contents related to one another?" Each of these has ad hoc engineering approaches that provide 40-80% solutions, and each of them begs for a deeply grounded formulation both to provide understanding and to attain the remaining 20-60% of practical value. The practical need is not purely technical but also depends on the user experience in application setup and configuration, and in collection and groundtruthing of sample documents. The challenge therefore extends beyond the science behind document image recognition and into user interface and user experience design.

show abstract

Markov Random Field Based Text Identification from Annotated Machine Printed Documents

Cited by 26 publications

References 13 publications

Connected Component Level Discrimination of Handwritten and Machine-Printed Text Using Eigenfaces

Connected Component Level Discrimination of Handwritten and Machine-Printed Text Using Eigenfaces

Overlapped text segmentation using Markov random field and aggregation

Scientific challenges underlying production document processing

Contact Info

Product

Resources

About