Abstract:

Euclidean distance is the most commonly used distance measurement method in the process of clustering analysis.The traditional Euclidean distance image segmentation method does not consider the spatial information, neighborhood characteristics and other factors. In order to use more image space information to improve the quality of image segmentation, in addition to implanting spatial constraints information of pixels, we propose an alternative neighborhood median weighted Euclidean distance to replace the Euclidean distance. The results of segmentation experiments on multiple images show that, compared with the existing algorithms, this algorithm cannot only improve  image segmentation effect with a better noise resistance, but also accelerate the convergence and obtain high efficiency.

Key words: clustering, Euclidean distance, image segmentation, neighborhood median weight, spatial constraints