Grey Clustering Analysis based performance evaluation on image segmentation

To acquire a comprehensive evaluation on image segmentation, this paper suggests an evaluation method based on Grey Clustering Analysis. In the method, several single-performance measures of segmentation results are discussed first, and then they are used to construct some whitenization weight functions and a clustering coefficient matrix. Next, on the basis of the matrix, the segmented images and their corresponding algorithms are clustered to three grades. Finally, according to the membership degree of the best grade, an overall evaluation result is gained. Our experiments show that the proposed method not only can cluster image segmentation algorithms or segmented images correctly, but also is efficient in automatic performance evaluation.