Language Identification in Degraded and Distorted Document Images

Abstract

Abstract. This paper presents a language identification technique that differentiates Latin-based languages in degraded and distorted document images. Different from the reported methods that transform word images through a character shape coding process, our method directly captures word shapes with the local extremum points and the horizontal intersection numbers, which are both tolerant of noise, character segmentation errors, and slight skew distortions. For each language studied, a word shape template and a word frequency template are firstly constructed based on the proposed word shape coding scheme. Identification is then accomplished based on Bray Curtis or Hamming distance between the word shape code of query images and the constructed word shape and frequency templates. Experiments show the average identification rate upon eight Latin-based languages reaches over 99%. . . .

Retrieval of machine-printed Latin documents through Word Shape Coding

“…In Ref. [17], we proposed to first remove impulse noise by a size filtering process where the filtering threshold depends heavily on the image resolution. Here, we instead suppress impulse noise by using a center-weighted median filter [19].…”

“…In this paper, we adopt the word shape coding scheme reported in our earlier work [17,18] and use it for the document vector construction. Two word shape features are utilized including the character extremum points and the number of horizontal word cuts illustrated in Fig.…”

“…Its number of horizontal word cuts 11 may be ambiguously interpreted as converted from extremum points over character descenders. Besides, instead of searching for the vector element one by one as done in [17,18], the element within a document vector is arranged in a descending order in term of the word frequency so that a newly converted word shape code may locate the matched vector element as soon as possible.…”

“…[17,18], we report a language filtering identification technique by using a word shape coding scheme, which converts each document image into an electronic document vector that captures the contents of image documents efficiently. In this paper, we adopt that word shape coding scheme and use it for document image retrieval.…”

Scripts and Numerals Identification From Printed Multilingual Document Images

“…[9][16] [1]. in such environment the large volume of data and variety of scripts makes such manual identification unworkable [9] [16]. In such cases the ability to automatically determine the script, and further, the language of a document, would reduce the time and cost of document handling.…”