Enhanced Edge Detection Algorithm Outperforms Traditional Canny Algorithm in Machine Learning Study

Strategic Insights -

Researchers at Al-Nahrain University have developed an improved edge detection algorithm by integrating machine learning into the traditional Canny algorithm. The new approach involves replacing the Gaussian filter with a bilateral filter and estimating Canny algorithm thresholds using the Flower Pollination algorithm. The enhanced algorithm was evaluated using 50 images from the Berkeley Computer Vision dataset, resulting in improved edge detection accuracy compared to the traditional Canny algorithm. According to the study, the enhanced algorithm has an AUC (Area Under the Curve) of 0.81 for the RF (Random Forest) classifier and 0.75 for the LR (Logistic Regression) classifier, outperforming the traditional Canny algorithm with an AUC of 0.57.

Key Takeaways:

  • The new edge detection algorithm proposed by Al-Nahrain University researchers replaces the Gaussian filter with a bilateral filter to improve performance.
  • The algorithm uses a new approach for estimating Canny algorithm thresholds through the Flower Pollination algorithm.
  • The enhanced algorithm was evaluated using 50 images from the Berkeley Computer Vision dataset.
  • The new algorithm has an AUC of 0.81 for the RF classifier and 0.75 for the LR classifier, outperforming the traditional Canny algorithm.
  • The improved algorithm was developed by researchers Russel Lafta and Zainab Sultani from Al-Nahrain University's Computer Science Department.
  • The study published in the Engineering and Technology Journal sets a new standard for edge detection performance.

Statistics:

  • 0.81: AUC for the RF (Random Forest) classifier using the enhanced algorithm.
  • 0.75: AUC for the LR (Logistic Regression) classifier using the enhanced algorithm.
  • 0.57: AUC for the traditional Canny algorithm.
  • 50: Number of images from the Berkeley Computer Vision dataset used to evaluate the enhanced algorithm.
  • 4: Issue number of the Engineering and Technology Journal where the study was published.
  • 43: Volume number of the Engineering and Technology Journal where the study was published.

Sources:

  • NewsRx. Al-Nahrain University Researchers Have Provided New Study Findings on Machine Learning (An optimize canny algorithm with traditional machine learning for edge detection enhancement). Journal of Engineering. May 19, 2025; p 108.
  • Engineering and Technology Journal. An optimize canny algorithm with traditional machine learning for edge detection enhancement. 2025, 43(4):253-259. (Available at: https://doi-org.sdpl.idm.oclc.org/10.30684/etj.2025.158177.1914)