19
TELUGU PALM LEAF CHARACTER RECOGNITION USING RADON TRANSFORM Presented by xxx

Palm leaf character recognition using radon transform

  • Upload
    -

  • View
    299

  • Download
    1

Embed Size (px)

Citation preview

Page 1: Palm leaf character recognition using radon transform

TELUGU PALM LEAF CHARACTER

RECOGNITION USING RADON TRANSFORM

Presented by –xxx

Page 2: Palm leaf character recognition using radon transform

Difficulties in the process of scanning and character recognition of Ancient palm leaves•Palm leaves lose flexibility

•Seepage of Ink

•Smearing along the cracks

•Deterioration of writing media

•Dirt and other discoloration

•Damage to the leaf

Page 3: Palm leaf character recognition using radon transform

2D Image acquisition

3D Image acquisition

•“DEPTH” proportional to pen pressure applied by the scriber can be gainfully exploited in PLCR

Page 4: Palm leaf character recognition using radon transform

IMAGE ENHANCEMENT

700 year old tulu-palm leaf before and after image processing

Page 5: Palm leaf character recognition using radon transform

Character Recognition

•Online character recognition

•Offline character recognition

For Indian languages,The recognItion accuracy for even handwritten paper documents is less than 60 %

Page 6: Palm leaf character recognition using radon transform

Challenges with Indian Languages•More number of composite characters

•OCR in English is highly developed. With smaller number of speakers, languages like Kannada/Telugu have not attracted equivalent efforts.

•In Indian scripts, consonants take modified shapes when attached to the vowels.

•Additionally, vertical extent of the character varies depending on the modifying vowel or consonant. Such characters are even more difficult for a machine to recognize.

•Non-uniformity in the spacing of the characters within a word due to the presence of conjuncts (vowel + consonant)

The presence of consonant conjuncts results in improper line segmentation. Recognition programs need to perform further processing to segment the lines.

In scripts like Devanagari, all the characters in a word are connected by a line called Shirorekha. Word separation and line separation is easy in these cases

Page 7: Palm leaf character recognition using radon transform

FUNCTIONAL PARTS OF A CHARACTER RECOGNITION SYSTEM

• DATA ACQUISITION

• BINARIZATION

• NORMALIZATION

• SEGMENTATION

• FEATURE EXTRACTION- extracting only those features that are of possible relevance to classification

Binarization

Page 8: Palm leaf character recognition using radon transform

RADON TRANSFORM

•transform two dimensional images with lines into a domain of possible line parameters• each line in the image will give a peak positioned at the corresponding line parameters.

Horizontal and vertical projections of radon transform Geometry of Radon transform

x' is the perpendicular distance of the beam from the origin θ is the angle of incidence of the beams.

Page 9: Palm leaf character recognition using radon transform

To represent an image, the radon function takes multiple , parallel beam projections of the image (beams spaced 1 pixel unit apart)from different angles by rotating the source around center of the image.

Single projection at a specified rotation angle

Sample of accumulator data of Radon transform

Page 10: Palm leaf character recognition using radon transform

FEATURE EXTRACTION

rescaling

Creating vector by concatenation operation

Page 11: Palm leaf character recognition using radon transform

Two different types of manuscripts selected for study-

Flow chart for data acquisition

Page 12: Palm leaf character recognition using radon transform
Page 13: Palm leaf character recognition using radon transform
Page 14: Palm leaf character recognition using radon transform
Page 15: Palm leaf character recognition using radon transform
Page 16: Palm leaf character recognition using radon transform
Page 17: Palm leaf character recognition using radon transform

Conclusions and future scope

1. Palm Leaf Character Recognition (PLCR) using Radon Transform is explored in this work.

2. Using only the (X-Y) co-ordinates the recognition accuracy for even the Hand written paper documents is less than 60% as reported in literature . The Palm leaf characters have additional information in the form of depth measured in microns, which is proportional to the stylus pressure applied by the scriber. Hence (X, Y, Z) information was used for the PLCR.

3. YZ plane images for the palm leaf characters gives good results (above 90% recognition accuracy), justifying the use of Z-dimension.

4. Although some of the characters like Va, Ma, Ya, Pa or Na are very similar to each other, the proposed is able to identify them distinctly in YZ- plane of projection. This shows the importance of Z-dimension, a special feature of palm leaf characters.

5. The Palm Leaf Character Recognition (PLCR) is performed only on basic Telugu characters and hence can be extended to Samyukt aksharas ( combination of 2 or more basic characters ).

6. The method of data collection can be improved in future, by automated process of measurements like a 3D laser scanner instead of manual data collection mainly for the Z- dimension.

Page 18: Palm leaf character recognition using radon transform

1. Senior and Robinson , “An Off-Line Cursive Handwriting Recognition System”, IEEE Transactions on Pattern analysis and Machine 2. Intelligence, Vol.20, No.3, 1998, pp. 309-321.2. Wakahara et al, “On-line handwriting recognition”, Special Issue of Proc. Of the IECC, Vol.80, No. 7, 1992, pp.1181-1194.3. Shi Zhixin, Setlur Srirangaraj and Govindaraju Venu. 2005. Digital Image Enhancement Using Normalization Techniques and their Application to Palm Leaf Manuscripts. CEDAR. Center For Excellence for Document Analysis and Recognition. New York. U.S.A.4. Ujjwal Bhattacharya and B.B.Chaudhuri, Handwritten numeral databases of Indian scripts and multistage recognition of mixed numerals,5. IEEE transcations on pattern analysis and machine intelligence, Vol.31 No.3, pp.444-457, March 2009.5.V.N.Manjunath Aradhya, G.Hemantha Kumar, S.Noushath,“Multilingual OCR system for South Indian Scripts and English documents: An approach based on Fourier transform and PCA”, Elsevier, Engineering applications of artificial intelligence, 2008, pp.658-668.6.S.N. Srihari, E.Cohen, J.J.Hull and L..Kaun, “A System to locate and Recognize ZIP Codes in Handwritten Addresses,” Int’l J. Research and Eng.-Postal Applications, Vol. 1, 1989, pp. 37-45.7. J.Tsukumo and H.Tanaka, “Classification of Hand printed ChineseCharacters Using Nonlinear Normalization Methods,” Proc. Ninth Int’l Conf. Pattern Recognition, 1988, pp. 168-171.8. A. Amin and H.B. Al-Sadoun, “Hand Printed Arabic Character Recognition System,” Proc. 12th Int’l Conf. Pattern Recognition, 1994, pp. 536-539.9 .H.Yamada, K.Yamamoto and T.Saito, “A Non-Linear Normalization Method for Hand printed Kanji Character Recognition—Line Density Equalization,” Pattern Recognition, Vol.23, 1990, pp.1023-1029.10. Panyam Narahari Sastry, Ramakrishnan Krishnan, Bhagavatula Venkata Sanker Ram, Telugu Character Recognition on Palm Leaves- A three dimensional Approach Technology Spectrum (JNTU Hyderabad), Vol. 2, No. 3, pp.19-26, November 2008.11. Panyam Narahari Sastry, Ramakrishnan Krishnan and Bhagavatula Venkata Sanker Ram, Classification and Identification of Telugu hand written characters extracted from palm leaves using decision tree approach, ARPN Journal of Engineering and Applied Sciences, Vol. 5, No. 3, March 2010.12.. Panyam Narahari Sastry, Ramakrishnan Krishnan and T.V.Rajinikanth “Palm leaf Telugu Character Recognition using Hough Transform” Proceedings of International Conference on Advanced Computing Methodologies (ICACM-2011), Elsevier Publication, December 2011, pp 21-28.

REFERENCES

Page 19: Palm leaf character recognition using radon transform

THANK YOU