원문정보
초록
영어
Influenced by light, houses, street lamps and other factors, tree images often contain noise and complex background information. The existing Otsu segmentation algorithms have deficiencies such as poor anti-noise capability, ignoring the class cohesion and so on. Based on the traditional gray value-neighborhood average gradient Otsu segmentation method, we propose an improved two-dimensional Otsu algorithm for tree image segmentation. The algorithm takes into account the between-class distance and within-class distance, which combined the average variance concept of two categories and proposed new threshold selection method, and reduce the interference of noise effectively. To achieve the best segmentation and reduce over-segmentation of background information, a method of removing small areas and morphological processing are used to optimize segmentation results. Experimental results show that the proposed algorithm has a good inhibition effect on noise and the effect of tree image segmentation is better than that of the traditional one.
목차
1. Introduction
2. 2D Otsu based on the Grayscale and Average Gradient
2.1. The Grayscale-Neighborhood Average Gradient Histogram
2.2. 2D Otsu Algorithm based on the Grayscale-Neighborhood Average Gradient
3. Improved 2D Otsu Threshold Segmentation Algorithm
4. Tree Image Segmentation Based on the Improved 2D Otsu Algorithm
5. Experimental Results and Analysis
6. Conclusions
Acknowledgements
References