A method that includes the fractal dimension to assess the similarity between retinal vessel segmentation images is presented. We propose that the method can contribute in the iterative refinement of vessel segmentation images. Manual vessel segmentation images are compared against an unsupervised automatic segmentation method. A small relative error is found when using standard metrics for comparison. The differences between segmentation methods are better understood by subdividing the images under study.
© 2018 The Author(s)PDF Article