Image Super-Resolution Based On Low-Rank Sparse Decomposition And Dictionary Learning

In the background of scientific and technological development and innovation,digital multimedia technology have been used widely in all areas of society.Among them,the image visual signal has become one of the main carriers of digital multimedia content transmission.A large number of imaging equipment(such as mobile phones,digital cameras and iPad)have been designed.Getting high-quality images has always been a goal that people are constantly pursuing.Due to the impact of various external uncertainties,people finally get the distorted image.After the researchers continue to study and explore,the image super-resolution reconstruction theory has been improved.Some practical results are being studied.Inspired by the application of compressed sensation in signal processing field,some effective reconstruction algorithms are thus proposed.Especially the image super-resolution reconstruction algorithm is proposed,which is attracted by researchers.Another theory related to the inherent theory of compression perception is the low-rank sparse decomposition theory of images.They are two different representations of data.Based on the combination of these two ideas,this paper presents a new image super-resolution reconstruction algorithm:1.After the analysis of image's structure,some of its microstructures are similar.Using its information characteristics,the image is decomposed in a certain way in this article.By measuring the euclidean distance between the image patches,select out the similar image patches.And they are quantified to form the image matrix.Due to the correlation between signal vectors,the matrix is a low rank matrix.According to the theoretical thought,in this paper,a new image reconstruction model is established to reconstruct the initial high resolution image,which can ensure that is consistent with the basic structure of the degraded low resolution image.2.The main task of this step is to restore the missing details of the initial high resolution estimate image.That is the missing high-frequency components in the image.The image reconstruction algorithm based on the sparse representation is one of the performance algorithm.This article draws on its ideas to achieve the purpose of this section.Especially different to the traditional dictionary training method,combining image matrices with low-rank sparse decomposition theory,this section makes innovations in the construction phase of the sample set.Not only improve the efficiency of dictionary training,but improve the quality of image reconstruction.In order to verify the effectiveness of the proposed algorithm,finally,designing the simulation experiments.Compared with some of the better performing reconstruction algorithms in this field,the evaluation criteria confirm that the proposed algorithm is more advantageous.The details of the image information recovery is better.