-on this paper, we proposed a new text line segmentation of handwritten and typewriting Arabic document images that uses the Outer Isothetic Cover (OIC) algorithm of a digital object. In the first step, we use this method to segment the composed document into text blocs. In the second step, for each text bloc we will extract the text lines. Finally, line text will be segmented into words or into pieces of Arabic word (PAWs). The first results obtained in the current stage of the proposed method over a dozen texts are encouraging. We have also tested this method on documents written in Latin scripts.
INTRODUCTION The first step in the automatic document recognition is the segmentation of the text image into text line. The objectives of this step is to assign each component of the text to the appropriate line; to make it possible to prepare the data for further processing such as normalization, word segmentation and feature extraction.
The segmentation of handwritten text is complicated by the variation of the distance interline and the undulation of baselines generate different orientations of the text. The characters in two lines of text may touch or overlap. This considerably complicates the segmentation line. In Arabic script, these situations are frequently due to the presence of ascending and descending characters. The massive presence of diacritical symbols often generates false lines.
Most work on the segmentation of a page in line is based on a decomposition of the image into connected components. After the separation of lines, we focus on the separation of words for each line, then segmenting each word in pieces (parts) of words.
In the framework of this article, we focus essentially on image segmentation of Arabic documents into blocks of text and lines. Then we apply our method to the segmentation of Latin documents.
In the following, we present some technique applied to the segmentation of documents into text lines. Then we present our approach to the segmentation of Arabic handwritings and printed documents into blocks, text lines and words or parts of words.
The results of our segmentation method are shown subsequently, by tests on historical and modern documents.
Finally we end our article with a conclusion and perspectives that show a possible extension of the proposed approach.
II.
Several works have been proposed for the segmentation
of documents. For example, Bennasri and al. have
proposed a method to extract lines of text Arabic script,
using the projection [
Nicolaou et al. [
There are also methods of segmentation using the
Hough transformed. This technique is widely used for
extraction of text lines [
Louloudis et al. proposed in [7], a technique for the extraction of lines and words of ancient Greek manuscripts. The Hough transform is applied to connected components using the centroids of rectangles encompassing their points as voters. These rectangles are estimated by calculating the average size of characters in the document. The proposed system was tested on the basis of documents ICDAR 2007 which is divided into 80 documents containing 1773 lines and 40 ancient manuscripts containing 1095 lines. The extraction rate of lines is 97%.
Another method used for segmentation is the snake or
the contour. With this technique Bukhari et al. proposed a
method for extracting parameterized snake lines of
handwritten documents [
Du et al. used the Mumford-Shah model [
A new method proposed by Vasant Manohar et al. in
[
III.
PROPOSED WORK
The proposed method realizes the segmentation of
handwritten and/or printed text lines, into words and into
pieces of words. It is based on the algorithm for the
construction of the isothetic covers of a digital object [
We thought to find a new segmentation method of document images. We started by construct the Outer Isethetic Cover (OIC) of documents. So, we made a change to this algorithm in order to segment a document into blocks of text, if the document is composed of text blocks, each block is then segmented into a set of lines then each line will be segmented into words and/or pieces of words.
A. Construction of the Outer Isothetic Cover (OIC)
In order to segment to segment document, we construct the outer isothetic covers of the corresponding document image after its binarization. To do this, we impose an isothetic set of grid size g on the binarized image.
Let Q1, Q2, Q3, and Q4 be the four quadrants incident at a grid point p(i, j), as shown in Fig. 1. The grid point p is decided to be a vertex depending on how many of the quadrants have object containment. Interestingly, there arise 24 = 16 different arrangements considering object containments of these four quadrants, which can be reduced to five cases. Let Cq (q = 0, 1, · · · , 4) denote the case of all the arrangements for which q out of 4 squares are occupied by the object. If p belongs to Case C1, then it is a 900 vertex of the isothetic polygon; and if it is a 2700 vertex, it belongs to C3. For Case C2, if the diagonal quadrants are occupied, then p is considered as a 2700 vertex; otherwise, p is a nonvertex grid point lying on some edge of the polygon.
For case C0, p is just an ordinary grid point lying
outside the polygon, whereas, for case C4, p is a grid point
lying inside the polygon [
So, in order to draw to draw polygon correspond of text
blocks, of text lines and of multiple words. We have
modified the algorithm TIPS [
First, the document is binarized. Then the grid points are traversed in the raw-major order until a 90° vertex (‘start vertex’) is found. Subsequent grid points are classified, marked as ‘visited’, and the direction is determined from each such grid point until the start vertex is reached. This completes the outer isothetic cover corresponding to an object (text blocks, text lines or word). The procedure is iterated over the remaining set of unvisited grid points until the next 90° vertex is found, which subsequently derives the polygon corresponding to another object. Finally, all the grid points are visited and the algorithm reports the vertex sequences of all the isothetic polygons corresponding to the text blocks, to text lines or words in the input document.
Setting the grid size g: In order that each isothetic polygon corresponds to each object and hence results in a sequence of vertices, specifying an appropriate grid size is necessary. So, for each case of segmentation, the grid size g is chosen after a set of tests on a set of document images.
A document can be composed by one or more paragraphs (or blocks of text). These text blocks can be arranged in parallel horizontally or vertically (Figure 2).
In this case, we thought to divide the document into parts of texts designed to simplify further processing i.e. segmentation of the text lines.
The extraction of these text blocks is made by varying the size of the grid g, the more the size g is large, the more between them. Contrary to the Latin script, a line can be segmented into word or into characters.
IV.
EXPREMENTS AND RESULTS In order to evaluate the results of our approach, we have the results are better. The figure 2 shows the results for document segmentation into blocks of text by changing the grid size g for handwritten Arabic texts.
A text or text block can be segmented into text lines. By changing the grid size g, we have applied our approach to segment a text into texts line. At the difference to segment a document to text blocks, the same algorithm is applied with the operator of mathematical morphology (the closure) in order to obtain the entire polygon line. The results are shown in Fig. 4.
D. Segmentation of line into words and/or into pieces of word :
Then we have used the algorithm to extract words from line. A line can be segmented into words and/or parts of word Arabic manuscript (printed respectively). This is because the Arabic writing is recursive. The word can be composed by parts of words (Pieces of Arabic Words (PAWs)) and sometimes there is enough space tested this algorithm on a variety of documents of two types of scripts: handwriting and printed. Handwritten Arabic documents composed by text blocks were segmented into blocks; the result is shown in Figure 1. As for the handwritten script, we have applied our method on documents printed Arabic and Latin;
The first database is a collection of 200 forms written by 200 different native writers. The writers were asked to write a paragraph of an Arabic text including up to 10 sentences. There were no restrictions for the writing. This database is an extension of the standard benchmarking IfN/ENIT database. The second collection of Arabic handwritten documents includes scans from historical documents collected during a research project in the IfN. The printed Latin text is from the Google Books (version 7, August 2007)
Then we tried to do the segmentation text blocks into text lines. The result of the segmentation of handwritten Arabic text is shown below. In the same way, we are showing in the following figures the result of such segmentation for printed Arabic and Latin script:
The last step of our method is to extract the connected components from the line. The result of handwritten Arabic text line segmentation into words or parts of words has been shown in Fig. 3. In the same step we are applied our method on a printed text line Arabic and Latin.
CONCLUSION
In this paper we presented some techniques of
segmentation methods. Then we proposed a new
segmentation method for document images handwritten
and printed script. The idea of this method is inspired
from the algorithm of construction of isothetic covers of a
digital object [
The results of our method are preferment for proper images document, especially for the type of printed texts document. This because in this type of script, the characters of two lines can neither touches nor overlaps. Instead of the handwritten, these situations exist frequently, which will sometimes give incorrect results.