Adaptive_Image_Segmentation_Based_on_PSNR_for_License_Plate_Recognition

https://www.pirahansiah.com/contents/publications/Papers/Adaptive_Image_Segmentation_Based_on_PSNR_for_License_Plate_Recognition

spotify

PDF Download My Conference Paper

Adaptive Image Segmentation Based on PSNR for License Plate Recognition

1. Introduction

• Objective: Propose an adaptive threshold method using Peak Signal-to-Noise Ratio (PSNR).
• Applications of Image Segmentation:
  • License Plate Recognition (LPR)
  • Preprocessing in image analysis
  • Object detection and classification

2. Pattern Recognition

• Applications:
  • Optical Character Recognition (OCR)
  • Biometrics (e.g., face, fingerprint recognition)
  • Medical diagnostics (e.g., X-ray analysis)
  • Military applications (e.g., Automated Target Recognition, Image Segmentation)

3. Motivation

• Thresholding:
  • Single and multi-level thresholding methods
  • PSNR-based threshold for improved accuracy in segmentation
• Challenges:
  • Differing environmental conditions (e.g., contrast, lighting)
  • Object complexity (e.g., fonts, colors, sizes)

4. Proposed Method

• PSNR as a Metric:
  • Used to determine optimal threshold value
  • Evaluated across various test cases (dark, medium, bright images)
• Algorithm:
  • Adaptive thresholding method steps:
    1. Calculate PSNR for different threshold values.
    2. Select the threshold value with the highest PSNR.
    3. Apply to image segmentation for license plate detection.

5. Multi-Level Thresholding

• Benefits:
  • More accurate segmentation in complex images.
  • Used when single-threshold methods are insufficient.
• Algorithm Steps:
  • Recursively apply PSNR to calculate the best multi-level thresholds.

6. Experimental Evaluation

• Comparison of Methods:
  • PSNR-based method vs. existing methods:
    • Kittler and Illingworth’s MET
    • Potential Difference
    • Otsu’s method
  • Results:
    • Proposed method: Reliable for high/low contrast situations (e.g., night, midday, rainy conditions).
• Performance Metrics:
  • Accuracy in detecting license plates and characters
  • PSNR values across different threshold levels

7. Conclusion

• Key Findings:
  • The proposed PSNR-based threshold method improves segmentation in license plate recognition.
  • Adaptive to changing environmental conditions (e.g., lighting, contrast).
  • Shows high reliability in experimental evaluations.