Research Of Image Segmentation Algorithm Based On Two-Dimensional OTSU’s Method

Image segmentation is one of the most important and critical research issues in image processing field and also has great influence on the subsequent tasks, e.g., image analysis, feature extraction and target recognition etc. So far, thousands of image segmentation algorithms have been proposed. However, thresholding segmentation methods are widely concerned by researchers due to its simpleness and effectiveness. Otsu’s method is a classic thresholding approach in image segmentation. While 2D Otsu’s method is shown to perform better than the original method in processing images corrupted by noise, it is found that this method is sensitive to Salt&Pepper noise and tends to generate incorrect threshold vector in this case. Besides, this method does not involve a step to deal with the pixels of noise and edges appropriately and therefore performs unsatisfactorily in case of low signal-to-noise ratio. In order to solve these problems, a novel robust image segmentation framework based on 2D OTSU called Median-Average 2D Otsu’s method is presented in this paper. It builds the 2D histogram based on the image smoothed by both median and average filtering, in contrast to the traditional method based on averaged image only. Then the optimal threshold vector is selected by two 1D searches on the 2D histogram. In addition, it introduces a region post-processing step to deal with the pixels of noise and edges.In this paper the research significance of image segmentation is introduced briefly and the image segmentation methods are summarized. Then we introduce the traditional 2D OTSU and analyze its problems. Finally, the proposed MAOTSU₂D is presented and the experimental validation of this method is reported. Experimental results on both synthetic and real images show that MAOTSU₂D improves the robustness to Salt&Pepper noise and Gaussian noise significantly compared with the traditional 2D OTSU and is also efficient.