Read and Recognition of old Kannada Stone Inscriptions Characters using MSDD Algorithm

Read and Recognition of old Kannada Stone Inscriptions Characters using MSDD Algorithm

Imran Khan, Ancy Elizabeth, Megha S G, Prakruti K S, Gowthami G R Department of ISE, GM Institute of Technology, Davangere, Visvesvaraya Technological University (VTU)

Abstract Invention of computer has made remarkable impact in all walks of life. Computer vision or Machine Vision has made our life simple and more dynamic. In the present work a successful attempt has been made using computer version technique to recognize and read the stone inscriptions belongs to ancient Kannada Characters. Conventional method is preparing such digital document take enormous amount of time, energy and money. The present approach provides an easy method of digitalization of stone inscriptions using simple camera. The techniques used here is simple and effective in removing noise and to recognize the characters in different condition like faded and unprotected old stone inscriptions.

In this work a novel approach called Mean, Standard deviation and Sum of Absolute difference Algorithm (MSDDA) is proposed Here in the present work, the Hoysala and Ganga periods characters are stored in a template. The character is identified based on its Mean, standard deviation and sum of absolute difference values. The corresponding recognized characters are converts into present Kannada script. An experimental result demonstrates high accuracy and better time efficiency in recognizing old Kannada stone inscriptions.

Keywords Mean, standard deviation and Sum of Absolute difference Algorithm (MSDDA), Mean value coefficient. Standard deviation

1. INTRODUCTION

   Kannada language has 2000 year of history and Karnataka is ruled by so many of dynasty. Ganga and Hoysala rulers were originally from Malnad Karnataka, an elevated region in the Western Ghats range. The Hoysala and Ganga era was an important period in the development of art, architecture, and religion in South India. The empires is remembered today primarily for its temple architecture. Over a hundred surviving temples are scattered across Karnataka, including the well-known Chennakesava Temple at Belur, the Hoysaleswara Temple at Halebidu, and the Kesava Temple at Somanathapura.

   In the present work we propose Mean, Standard Deviation and Sum of Absolute difference Algorithm (MSDDA) to recognize Hoysala and Ganga periods characters. In this work, the database of Hoysala and Ganga periods stone inscriptions characters are stored in .jpg and .wave format. The user can choose any characters in the test image. When they select at that time, it segment that characters and performs pre- processing and it extract features from those characters. Depending on extracted feature the algorithm calculates mean and standard deviation values of corresponding characters and finally it compare that calculated value of the same with database characters and it recognizes the corresponding characters.

2. PROPOSED ALGORITHM

   The Algorithm flow is shown in figure.l

   1. Database creation of Ganga and Hoysala phase Kannada characters.

   2. Test image:

   Here first capture Kannada Stone inscriptions images using normal digital camera

   c Characters selection:

   In this user can select any four characters simultaneously in test image.

   1. Pre-processing

      Selected characters are pre-processed and feature extraction can be done.

   2. Mean standard deviation and Sum of Absolute difference Algorithm (MSDDA).

   In this step, the preprocessed characters will be recognize based on its feature extraction and mean, standard deviation averaging values

   Figure 1: Flow chart of present work.

3. METHODOLOGY AND IMPLEMENTATION

   1. Old Kannada Stone inscriptions

      In this proposal, 12° century Hoysala and Ganga periods Stone inscriptions in Doddagavanahalli near 10KM from Hassan are captured. More than 50 images of old Kannada stone inscriptions are captured and selected only 2 exigent Hoysala, Ganga stone inscriptions for better recognition. Selected captured Hoysala and Ganga period stone inscriptions are shown below Figure 2a & 2b.

   2. Pre-processing

      1. Gaussian filtering

         Gaussian filtering is used to blur images and remove noise and detail. In one dimension, the Gaussian function is as G(x)=(1/(Sqrt(2pie * sigma)))e^ (x^2/sigma^2)

         Where is the standard deviation of the distribution. We have Also assume that the distribution has a mean of zero ( i.e. it is centered on the line x=0).

      2. Edge detection

         Here Sobel edge detection technique is used and Sobel calculates maximum gradient of image and it perform smoothing of image ,if image is smooth then the information found in that can easily extract.

      3. Opening

         The edge detected image is very thin so dilation followed by erosion is a perform called opening. It increases the thickness of edge detected image and dilation combines two similar pixels like AUB and erosion performs ABB.

      4. Characters reconstruction

      The opened image contains some breaking border and that border is reconstructed by filling of holes. Reconstruction is a process of constructing an image by filling pixel into broken border and this reconstruction is very important for character recognition.

   3. Mean standard deviation and Sum of Absolute difference Algorithm (MSDDA).

   1. Database: Store selected old and current alphabets in JPG format with each image having size 24*42 dimensions.

   2. Image mean value: Calculate mean value of database and test images. Mean calculates the average pixel value of each image.

   3. Standard deviation. Standard deviation is a square of some co-efficient. This is a second computation process in image. Thestandard deviation (SD) measures theamount of variation or dispersion from the average.

      Let X be a random variable with mean value

      SD= = BE [X-p] (2)

   4. Averaging

      In this it calculates average of mean and standard deviation and for accurate recognition averaging of Mean and standard deviation is performed.

      AVG = (MV+SD)/2

   5. Sum of absolute difference

   (3)

   Absolute difference measures the two difference points. Here in this it calculates test image and database images averaged difference value.

   Figure 5.

   When

   SAD= AVG Test image AVG Database image

   MV=Mean Value SD=Standard deviation SAD=Sum of Absolute Value AVG= Average

4. RESULTS

   (4)

   l) Create database of Hoysala and Ganga period.

   2) Test image

   Capture some old stone inscriptions characters belong to Hoysala and Ganga periods using 16 Mega pixel ordinary digital cameras as shown in figure 3.

   Figure 5: pre-processed image

   5) Mean, standard deviation and Sum of Absolute difference Algorithm (MSDDA)

   In this calculation of mean, standard deviation value and sum of absolute difference value of image takes place and overall calculation and matching is performed by MSDDA algorithm and here the corresponding character is recognize on the basis of mean, standard deviation and sum of absolute value.

   In table 1 show Mean, Standard deviation and absolute difference value of database images and test image as shown in table 1.

   Figure 3: test images

   1. Characters selection

      In this step, characters will be select by user on their requirement as shown in figure 4

      When

      SAD=Avg(MV, SDyg9 MV, SD of Data base images) (8 )

      SAD= Sum of Absolute Deerence SD=Standard deviation value MV= Mean Value

      Likewise it read and recognized all characters shown in

      figure 6

      Figure 4: Characters selection in Stone inscriptions

      In this proposal Ga, Ja, ya and Tha, characters of 12° century stone inscriptions are selected.

   2. Pre-processing

   Figure 6: Read and Recognition of stone inscriptions Hoysala characters

   TABLE 1: Mean and standard deviation and Sum of Absolute Deference values of test and database images

   Images	Mean Value (M(	Standard deviation value (SD)	Sum of Absolute Deference (SAD)	Recognition Status
   	173.05	121.0578	0	Recognized
   	177.32	112.4213	4.2691	Not Recognized
   	164.07	115.64	40.4788	Not Recognized
   	163.05	121.9352	17.8327	Not Recognized
   	132.57	122.0158	1.5532	Not Recognized
   			8.0918	Not Recognized
   	171.49	112.6063	13.9538	Not Recognized
   	181.14	119.3058	44.6135	Not Recognized
   	159 .09	121.0578	1.6958	Not Recognized
   		114.8631	8.9729	Not Recognized

   The Recognition rate is calculated using equation 6 and achieved rate is shown in Table 3, 4.

   The Recognition rate is obtain by formula

   When p Recognition rate, m= accurate match, 7 = inaccurate match, Q = Number of samples

   (6)

   TABLE 2: Hoysala Period stone inscription characters Recognition rate analysis

   Test images	Recognized Character
   		10	0	100%
   		10	0	100%
   		19	i	95%
   		20	0	100%
   		20	0	100%
   		20	0	100%
   		20	0	100%

   Test images		Recognized Character
   			20	0	100%
   			20	0	100%
   			19		95%
   			20		100%
   			20	0	100%
   			19		95%
   			20	0	100%

   Test images		Recognized Character
   			20	0	100%
   			20	0	100%
   			19		95%
   			20		100%
   			20	0	100%
   			19		95%
   			20	0	100%

   TABLE 3: Ganga Period stone inscription characters Recognition rate analysis

   From above Table 2 and 3 the recognition rate is 98.75 % accuracy for Hoysala and Ganga stone inscriptions characters.

5. CONCLUSIONS

In this present work a novel method of old Kannada character recognition is developed. It is easy and useful Recognition method for to identify old Kannada characters. Here in this the digitalization of old Kannada inscriptions was performed and it helps a common man to knowing the present Kannada literature. The software developed has been tested with chosen through sixteen characters with 40 data sets. The result achieved is efficient with 98.75% accuracy.

REFERENCES:

1. David G Distinctive Image Features from Scale-Invariant Keypoints, Computer Science Department University of British Columbia Vancouver, B.C., Canada International Journal of Computer Vision, 2004.

2. Kumar, V. Devadoss Ambeth, and M. Ramakrishnan. "A comparative study of fuzzy evolutionary techniques for footprint recognition and performance improvement using wavelet-based fuzzy neural network", International Journal of Computer Applications in Technology, 2013.

3. M. V.Jeya Greeba, G.Bhuvaneswari Recognition of Ancient Tamil Characters in Stone Inscription Using Improved Feature Extraction, international Journal of Recent Development in Engineering and Technology, Volume 2, Special Issue 3,February 2014.

4. Rajakumar S, Dr.Subbiah Bharathi V 7" century ancient Tamil Character recognition from Temple wall inscriptions Indian Journal of Computer Science and Engineering (IJCSE), Vol. 3 No.5 Oct-Nov 2012.

5. Ankit Sharma English Character Recognition Using Neural Network, International Journal of Engineering Trends and Technology (IJETT), Volume4Issue4-April 2013.

6. Faisal Mohammad.et.al, Optical Character Recognition Implementation Using Pattern Matching, International Journal of Computer Science and Information Technologies, Vol.5 (2), 2014, 2088-209.

7. Kirill Safronov .et.al, Optical Character Recognition Using Optimisation Algorithms, Proceedings of the 9t ' International Workshop on Computer Science and Information Technologies CSIT2007, Ufa, Russia, 2007

8. Rahul KALA Offline Handwriting Recognition using Genetic Algorithm, IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 2, No I, March 2010.