An Experiment with Kernel Graph Cut and GMM Based Hidden Markov Random Field Image Segmentation Techniques
Anuja Deshpande · International Journal for Research in Applied Science and Engineering Technology · 2018
This paper discusses segmentation outcome of experiments conducted on Kernel Graph Cut and Gaussian Mixture Model based image segmentation techniques. The main objective of this experiment is to understand Effectiveness of these segmentation techniques on specific natural images having complex image composition. Effectiveness is assessed using human visual assessment and mathematical models such as Jaccard Index, Dice Coefficient and Hausdorrf Distance by comparing the segmented images with ground truth. While both techniques employ k-means as the clustering algorithm, this experiment findings suggest GMM based technique to be better over Kernel Graph Cut in terms of completeness and overall quality of segmentation.