An improved adaptive image segmentation algorithm based on edge detection
MEI Fei1, PENG Huilin 1, MEI Chao 2, LI Junyan 3,TONG Yala 1
(1.College of Science, Hubei University of Technology, Wuhan 430079, China;2.College of Civil Engineering, Architecture and Environment, Hubei University of Technology, Wuhan 430079, China;3. School of Computer Science, Hubei University of Education, Wuhan 430205, China)
Abstract:The balance between efficiency and accuracy has always been a challenge in image segmentation tasks, especially under variable background conditions. The traditional MageFreehome algorithm is often difficult to meet the requirements of efficiency and accuracy at the same time when dealing with such problems. To solve this problem, an improved scheme for image adaptive segmentation based on edge detection has been proposed in this paper. Firstly, the OTSU algorithm is used to find the optimal dynamic threshold, so that the image segmentation process can adaptively adjust the threshold to adapt to the image under different background conditions. Then, a new method for selecting the center point of the image is proposed to solve the defect that the center of the image falls outside the target area of segmentation. Finally, a method for determining the start/end point of the segmentation region based on continuity detection was designed, which further improved the accuracy of image segmentation. The verification of image use effect shows that the three strategies in this paper have a positive impact on the accuracy of image segmentation. The experimental results of the improved algorithm with the MagFreehome algorithm on different orders of magnitude show that the accuracy and recall rate of the improved algorithm are significantly improved while maintaining high efficiency, and the processing speed remains almost the same. Ablation experiments further confirmed the effectiveness of the combination of different strategies, and strategy 1 and strategy 3 were particularly outstanding in improving accuracy and recall, bringing significant improvements of 7.29% and 5.71%, respectively, and when the three strategies were applied at the same time, the accuracy and recall rate increased by 15.51% and 16.72%, respectively. The improved scheme in this paper effectively balances the problem of image segmentation efficiency and accuracy, and provides an effective solution for application scenarios that require fast and accurate image segmentation, such as multimedia teaching, autonomous driving, etc.