PEAK_SIGNAL-TO-NOISE_RATIO_BASED_ON_THRESHOLD_METHOD_FOR_IMAGE_SEGMENTATION

https://www.pirahansiah.com/farshid/portfolio/publications/Journals/PEAK_SIGNAL-TO-NOISE_RATIO_BASED_ON_THRESHOLD_METHOD_FOR_IMAGE_SEGMENTATION/

spotify

PDF Download

Mind Map: Peak Signal-to-Noise Ratio Based on Threshold Method for Image Segmentation

1. Introduction

• Importance of Thresholding: Separates objects from the background, crucial in image processing and pattern recognition.
• Types of Thresholding:
  • Single Thresholding: Produces binary images (0 and 1).
  • Multilevel Thresholding: Produces images with pixel values between 0 and 255.
• Objective: Develop a new algorithm using Peak Signal-to-Noise Ratio (PSNR) for image segmentation.

2. Thresholding Techniques

2.1 Single Thresholding

• Definition: Uses a single threshold value to convert the image into binary.
• Goal: Maximize segmentation accuracy and reduce storage requirements.
• Methods:
  • Kittler and Illingworth’s MET.
  • Potential Difference.

2.2 Multilevel Thresholding

• Definition: Uses more than one threshold value to segment images.
• Goal: Applied when a single threshold is insufficient for global segmentation.
• Method:
  • Arora’s Recursive Algorithm.

2.3 Multi-Thresholding

• Definition: Applies multiple threshold values, calculates blobs/objects in the image.
• Goal: Selects peak threshold values based on object count.
• Method: Abdullah’s Multi-Threshold Algorithm.

2.4 Otsu’s Method

• Definition: Unsupervised, nonparametric method for automatic threshold selection.
• Goal: Maximizes the variance between classes to select optimal thresholds.

3. PSNR-Based Thresholding

• Proposed Method: Utilizes PSNR as a measure to select the optimal threshold.

• Formula:PSNR = 10 * log10(MAX^2 / MSE) where MSE is the mean squared error and MAX is the maximum possible pixel value.

• Algorithm:
  1. Calculate PSNR for each threshold value.
  2. Select the threshold that maximizes PSNR.

4. Results and Evaluation

• Datasets:
• DIBCO 2011 (handwritten and printed images).
• Metrics:
• F-Measure: Evaluates binary classification accuracy.
• PSNR: Measures similarity between the original and thresholded images.

4.1 Performance

• The proposed method showed better PSNR and acceptable F-measure for printed images.
• Comparison with Kittler and Illingworth, Otsu, and other methods highlighted superior performance of the proposed method in certain conditions.

5. Application Areas

• Automated Visual Inspection.
• License Plate Recognition.
• Handwritten Image Segmentation.
• Image Processing in Mobile Devices (with resource limitations).

6. Conclusion

• Proposed Method: Adaptive PSNR-based thresholding is a viable approach for image segmentation.
• Results: Competitive performance in standard, printed, and handwritten images.
• Future Work: Potential for further improvement, especially in handling complex images with varying contrast.