Image Super-resolution Reconstruction Utilizing the Combined Method of K-SVD and RAMP

Abstract

A new image super-resolution reconstruction (SRR) method, combined a modified K-means based Singular Value Decomposition (K-SVD) and Regularized Adaptive Matching Pursuit (RAMP) algorithm is proposed in this paper. In the modified K-SVD algorithm, the maximum sparsity is considered. First, the K-SVD denoising model is first to preprocess the Low Resolution (LR) images. And then, the high-resolution (HR) and LR images are both trained by the RAMP based K-SVD algorithm. The LR and HR dictionaries are also classed by the K-mean method. In test, a human-made and real LR image, namely millimeter wave (MMW) image are respectively used to testify our method proposed. Further, compared our image SRR method with methods of the basic K-SVD and RAMP, experimental results testified the validity of our method proposed.KeywordsImage super-resolution reconstructionSparse representationHigh resolution (HR)Low resolution (LR)K- Singular Value Decomposition (KSVD)Regularized adaptive matching pursuit (RAMP)