Vous êtes sur la page 1sur 4

Character Recognition Using Chain Code

Dipali Gawade, Sheetal Pawar College Of Engineering, Pune. pawarsheetal13@gmail.com

Abstract: This paper presents the


simplified chain code from our concept of basic shape to recognize the characters. The proposed chain code is based on the eightdirection Freeman chain code. Hence we focus on Shape of the contours which characterize the recognized alphabet. Index Terms: Character Recognition, Simplified Chain Code, Basic Shape.

Introduction: Recognition of handwritten


characters has been a popular area of research for many years and still remains an open problem. It has a versatile range of application domain, including postal automation, bank check processing, automating of processing of large volumes of data, language based learning, ledgering catalogue for library and reading aid for blind. The proposed work is an attempt for offline handwritten character recognition problem by concentrating mainly on chain code. Chain codes have been claimed as one of the techniques that are able to recognize characters and digits successfully. This is because of several advantages possessed by this technique. The first advantage over the representation of a binary object is that the chain codes are a compact representation of a binary object. Second, the chain codes are a translation invariant representation of a

binary object. Due to that, it is easier to compare objects using this technique. The third advantage is that the chain code is a complete representation of an object or curve. This means that we can compute any shape feature from the chain codes. Chain codes provide a lossless compressing and preserving all topological and morphological information which bring out another benefit in terms of speed and effectiveness for the analysis of line patterns. It is also suitable For detecting feature of a region such as sharp corners, area, perimeter, moments, centers, eccentricity, projection and straightline segments. Chain codes are invertible hence object can be reconstructed back from its chain code representation.

Chain Code Schemes:


First chain code scheme was introduced by Freeman in 1961 that is known as Freeman Chain Code. The basic idea behind the chain code is that each boundary pixel of an object has an adjacent boundary pixel neighbor whose direction from the given boundary pixel can be specified by a unique number between 0 and 7 (8-connectivity neighborhood).Chain codes could also be defined using 4-connectivity neighborhood . In the following discussion we use 8connectivity neighborhood. Given a pixel, consider its 8 neighboring pixels. Each 8 neighbor can be assigned a number from 0-7 representing one of the 8 possible directions from the given pixel (see Fig.1).This is done using the same orientations throughout the entire image.

direction of chain code segments; the numbers of left hand turns of /2 or /4 needed to achieve the direction of next chain segment.

Chain code properties:


1. Chain codes describe an object by a sequence of unit- size line segment6 with a given orientation. 2. The first element of such a sequence must bear information about its position to allow reconstruction of the region. 3. Even codes {0,2,4,6} corresponds to horizontal and vertical directions ;odd codes {1,3,5,7} correspond to diagonal direction. 4. Each code can be considered as the angular direction, in multiples of 45 degrees that we must move to go from one contour pixel to the next. 5. The absolute co-ordinates of the first contour pixel (e.g. top, leftmost) together with the chain code of the contour represent a complete description of discrete region contour. 6. When there is a change between the two consecutive chain codes, then the contour has changed the direction. This point is defined as a corner.

Fig.1

The chain code for the boundary of a binary image is a sequence of integers c= { , , .... } having each from the set {0,1,2, ......,7 } for i=0.1,....,n-1. The number of the elements in the sequence c is called the Length of the chain code. Chain codes may be made position-independent by ignoring the starting point. If they represent the close boundaries they may be Start point normalized by choosing the start point so that the resulting sequence of direction codes forms an integer of minimum magnitude .

Methodology
The Fig.3 below shows the methodology. There are six phases involved; image acquisition, data definition, image preprocessing, image segmentation, chain code derivation and character recognition. The images of car plates are varied in terms of its sizes and fonts for the testing process (during recognition).

Fig.2 : Chain code in 4-connectivity and its derivative. Code : 3,0,0,3,0,1,1,2,1,2,3,2 Derivative: 1.0,3,1,1,0,1,3,1,1,3,1 The derivative of the chain code is useful because it is invariant under the boundary rotation. The derivative is simply another sequence of numbers indicating the relative

Pre-processing:
Two processes are involved in preprocessing which are threshold and filtering. Images used are gray-scaled images and are converted into binary images which means that every pixel in the image is convert to the binary values (0 and 1). Fig.3 : The proposed Methodology

Segmentation:
Two processes have been done in this phase; the boundary extraction and segmentation. The boundary extraction of character is done in order to ease the process of deriving the chain codes. The segmentation has been done using the pixel count technique.

Data Definition:
Data that comes from the form of images need to be identified and analyzed first before tested. The information about types of fonts used, need to be gathered. Based on the observation on collected images, the various styles of writing can be recognized forming as raw data.

Chain Code Derivation:


This phase is to derive the chain codes for character in the specific region which is the result from image segmentation phase. The algorithm for extracting chain codes for 8connected boundaries : 1.Find the pixel in the object that has the leftmost value in the topmost row, call this pixel P0. 2. Define a variable dir (for direction), and set it to equal to 7(since P0 is the top-left pixel in the object, the direction to the next pixel must be 7). 3. Traverse the neighborhood of the current pixel in a counter-clockwise direction, and code the direction using 3-bit binary number. 4. Stop when the current boundary element Pn is equal to first boundary element P0.

Recognition:
The character recognition has been done by using chain codes derived for character from the previous phase. The code is then used as a guide to recognize the characters by matching the chain codes extracted from the previous phase.

Conclusion:
Chain code is one of the most interesting topics in pattern recognition and image processing. Since introduced by Freeman in 1961, chain code increases rapidly. There are many extended and applications of the chain code time by the time Most of the new chain codes are extended from Freeman chain code and Vertex chain code that introduced by Bribiesca. The new chain codes need the binary image as the input, because chain code is the code based on boundary. Chain code could be applied in many parts of pattern recognition and image processing. Then it is important to extend some algorithm about the chain code, especially vertex chain code.

[5] Tran Duc Duan, Tran Le Hong Du, Tran Vinh Phuoc, Nguyen Viet Hoang, Building an Automatic Vehicle License-Plate Recognition System, 3rd International Conference on Computer Science Research, Innovation & Vision for the Future (RIVF05), pp. 59-63, 2005.

References:
[1] Xu Jianfeng, Li Shaofa and Chen Zhibin (2003). Color Analysis for Chinese Car Plate Recognition. Proceedings of the IEEE International Conference on Robotics, Intelligent Systems and Signal Processing. pp1312 1316. [2] Pedro Barroso, Joaquim Amaral, Andr Mora, Jos Manuel Fonseca, Adolfo SteigerGaro (2004). A Quadtree Based Vehicles Recognition System. 4th WSEAS International Conference on Optics, Photonics, Lasers And Imaging (ICOPLI 2004). pp12-16. [3] Wenjing Jia, Huaifeng Zhang and Xiangjian He, Region- Based License Plate Detection, Journal of network and Computer Applications, Elsevier, In Press, Corrected Proof, Available online 17 November 2006. [4] Yo-Ping Huang, Shi-Yong Lai, Wei-Po Chuang, A Template-Based Model For License Plate Recognition, Proceedings of the 2004 IEEE International Conference on Networking, Sensing and Control, vol. 2, pp. 737 742, 2004.

Vous aimerez peut-être aussi