This paper presents a novel image thresholding algorithm, namely Random spatial sampling and Majority voting based Image Thresholding (RMIT) algorithm. The proposed image thresholding algorithm RMIT firstly obtains a population of thresholded sub-images by using random spatial sampling and the well-known Otsu's image thresholding algorithm, then aggregates all obtained binary sub-images into a consensus binary image via majority voting. Since the sub-images are randomly selected with different sizes ranging from one pixel to the entire image, RMIT can make use of both global and local information for thresholding an image without any prior knowledge about the image. The effectiveness of RMIT is confirmed by experimental results on benchmark real images. © 2010 IEEE.
|Title of host publication||2010 International Conference on Machine Learning and Cybernetics, ICMLC 2010|
|Publication status||Published - 2010|
- Image thresholding
- Majority voting
- Random spatial sampling