GrabCut and support vector machine based robust color image segmentation and classification model

Abstract

Image segmentation is a way of dividing an image to useful disjoint portions where every portion is almost similar with no
overlapping. Color image segmentation depends on the color features of the image pixels considering that the identical
color in the image belongs to individual cluster and thus meaningful objects in the image. It is a fundamental challenge in
almost all types of image processing and computer vision applications like object identification, image interpretation,
medical imaging, etc. These image based applications are highly based on image segmentation that identifies the quality of
the image investigation and understanding. In this paper, we develop robust image segmentation with classification
technique. The proposed model operates on two stages namely image segmentation using automatic modified Grab cut
(MGC) technique and image classification using support vector machine (SVM). A detailed experimental analysis is made
using a set of benchmark color images against state of art approaches. A clear qualitative analysis is made to neatly identify
the competence of the presented model. The presented MGC method attains a lowest average error rate of 8.175 whereas
the neural network (NN) achieves a higher average error rate of 15.796.The attained experimental values ensured the
superiority of the proposed model over the compared methods