Image inpainting with LS-SVM based on additive high order kernel

Abstract

This paper discusses the application of least squares support vector machine (LS-SVM) in image inpainting. The data with strong correlation with the damaged area are selected to train the LS-SVM model, and then predict the damaged parts with the obtained model. In order to make full use of the correlation in the image, this paper employs the additive high order kernel function to improve the prediction accuracy of LS-SVM model. The experimental results show that the presented LS-SVM model improves the inpainting obviously.