Weight selection in W-K‐means algorithm with an application in color image segmentation

In this paper, a weight selection procedure in the $W$ ‐ $k$ ‐means algorithm is proposed based on the statistical variation viewpoint. This approach can solve the $W$ ‐ $k$ ‐means algorithm’s problem that the clustering quality is greatly affected by the initial value of weight. After the statistics of data, the weights of data are designed to provide more information for the character of $W$ ‐ $k$ ‐means algorithm so as to improve the precision. Furthermore, the corresponding computational complexity is analyzed as well. We compare the clustering results of the $W$ ‐ $k$ ‐means algorithm with the different initialization methods. Results from color image segmentation illustrate that the proposed procedure produces better segmentation than the random initialization according to Liu and Yang’s (1994) evaluation function.