Eigenspace method for text retrieval in historical document images

Abstract. This paper presents an automatic keyword extraction method from historical document images. The proposed method is language independent because it is purely appearance based, where neither lexical information nor any other statistical language models are required. Moreover, since it does not need word segmentation, it can be applied to Eastern languages where they do not put clear spacing between words. The first half of the paper describes the algorithm to retrieve document image regions which have similar appearance to the given query image. The algorithm was evaluated in recall-precision manner, and showed its performance of over 80-90% average precision. The second half of the paper describes the keyword extraction method which works even if no query word is explicitly specified. Since the computational cost was reduced by the efficient pruning techniques, the system could extract keywords successfully from relatively large documents.

show abstract

“…3. This method is similar to our previous study [9], but differing in some respects because it includes our recent improvement.…”

Section: Ranked List Makingmentioning

confidence: 96%

“…The proposed method starts with making ranked list sorted by similarity to a query image, as described in our previous study [9]. Section 2 reviews the method for making ranked list, and besides, its performance is evaluated in recallprecision manner.…”

Section: Introductionmentioning

confidence: 99%

Automatic Keyword Extraction from Historical Document Images

Terasawa

Nagasaki

Kawashima

2006

Document Analysis Systems VII

View full text Add to dashboard Cite

show abstract

“…The purpose of their worls is to extract and recognize words using description based on computed features. Terasawa et al [11] also performed word spotting inspired by an Eigen space method [11] or gradient [12]. In this case, word signatures are extracted from sliding windows.…”

Section: Related Workmentioning

confidence: 99%

Local Orientation Extraction for Wordspotting in Syriac Manuscripts

Bilane

Bres

Emptoz

2008

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. This paper presents a contribution to Word Spotting applied for digitized Syriac manuscripts. The Syriac language was wrongfully accused of being a dead language and has been set aside by the domain of handwriting recognition. Yet it is a very fascinating handwriting that combines the word structure and calligraphy of the Arabic handwriting with the particularity of being intentionally written tilted by an angle of approximately 45°. For the spotting process, we developed a method that should find all occurrences of a certain query word image, based on a selective sliding window technique, from which we extract directional features and afterwards perform a matching using Euclidean distance correspondence between features. The proposed method does not require any prior information, and does not depend of a word to character segmentation algorithm which would be extremely complex to realize due to the tilted nature of the handwriting.

show abstract

“…Fink and Plötz [7] tested appearance-based features and compared it with heuristic features. Terasawa et al [8] developed principal component analysis-based descriptors and they also developed gradient distribution features [9] for word spotting for Japanese handwritten documents.…”

Section: Related Workmentioning

confidence: 99%