Digit-recognition-based-on-using-template-matching-Presentation.pdf

Digit recognition based on using template matching 1673_B2_ANIK DEB NATH PREMIER UNIVERSITY

1. Digit recognition based on
using template matching
Presented By:
Anik Deb Nath
ID: 1803510201673
Dept. of CSE, PUC

2. Contents
Motivation
Objective01
02
Applications03
Methodology04
Processing Example05
Conclusion06
Appendix07

3. g
(b)
01
Objectives
•Recognition digit for given image using template matching .
Figure for Processing Example : (a) Template & candidate image (b)
Template matching algoithm (c) Recognize digit image
Matching similarity by NCC
(a)
(c)

4. 02
Motivation
Traditional methods often struggle to handle various fonts, styles, and
handwriting.01
02
03
Ensuring accurate digit extraction and interpretation,
To minimize human error, inefficiencies and inaccuracies.

5. Applications
•Numeric entries in forms filled up by hand
•Processing bank check amounts
•Room Number
03

6. Methodology 04
Input image normalize by size
Convert images to grayscale.
Convert to binary using Otsu's thresholds
Complement Binary Images
Calculate Candidate Mean
NCC Loop
Find Best Match
Visualize result by rectangle box
Template
Candidate(digit)

7. Process Example - 105
Fig 1: Read RGB image and convert
to grayscale image
Fig 2: Convert gray to binary and
take the complement of binary
image
Fig 3: Display the current template,
candidate and best match

8. Process Example -206
Fig 1: Read RGB image and convert
to grayscale image
Fig 2: Convert gray to binary and
take the complement of binary
image
Fig 3: Display the current template,
candidate and best match

9. Conclusion
07
In conclusion, the code demonstrates a basic implementation of digit recognition using the NCC
technique.
Merits of NCC:
•Easy to understand and implement.
•Effective for finding patterns.
•Clear interpretation of results.
•Doesn't need extensive training.
Demerits of NCC:
•Can be slow for large images.
•Only good for exact or near-exact matches.
•Affected by similar patterns in the background.
•Challenging to set accurate thresholds.
•Struggles with noise, and variations.

10. Appendix
08

11. Appendix
09

12. Appendix
10

13. Appendix
11

14. Thank You