Paper Title
Application of Kernel Regression in Single Image Super Resolution
Abstract
Super resolution is an enhancement technique used for converting one or more low quality image in high quality
image. Single image is effectively resolved with two methods interpolation based and learning based method. Traditional
interpolation methods are unable to produce sharp edges and clear details. So learning based method is applied in this paper.
Second order kernel regression is used at the output of testing and training phase to reduce the mapping error. Training
features are extracted using K-singular value decomposition dictionary. Kernel allow data to map into high dimensional for
increasing computational efficiency. This method is capable to reduce use of dictionary. Comparison between old method
and presented method were done based on peak signal to noise ratio and mean square error values. Result of experimentation
shows that implemented SR method is more efficient and robust to different data types. Result shows that implemented SR
method has much future scope with high resolution factor.
Keywords - Kernel regression, K-singular value decomposition, steering matrix