The Research Of Image Denoising Algorithm Based On Compressed Sensing

Image is one of the important ways to get information, so it is an important content toprocess image in information theory. While images will be contaminated by the noise in theacquisition, transmission and storage process, resulting in blurred images. And therefore it isvery important to remove the noise to acquire high quality images. In this paper we make a studyof a new theory called Compressed Sensing, include its application on the area of imagedenoising. In this theory, Compressive Sensing(CS) is a novel signal sampling theory under thecondition that the signal is sparse or compressible, using an non-adaptive linear projection tokeep the structure of the original signal, and reconstructing the original signal accurately throughsolving the optimal problem. Compressive sensing breaks the bottleneck of Shannon theorem,and saves space and cost in greatly in the process of data transmission, been called the newbreakthrough in the data collection technology.This paper discusses the CS theory concretely and then introduces the application in thearea of image reconstruction. In the next, some familiar signal sparse decomposition algorithmsinclude their principles and steps are researched deeply, such as Matching Pursuit algorithm(MP),Orthogonal Matching Pursuit algorithm(OMP). Then briefly introduce the core ideas of thesingular value decomposition algorithm and k-means algorithm, discuss the relationship betweenk-means algorithm and K-SVD algorithm, expound the principle and advantages of K-SVDalgorithm. Through the simulation, the differences of denoising effect are obvious by comparingthe wavelet threshold denoising algorithm and K-SVD algorithm. Based on lots of experiments,we can summarize K-SVD algorithm is a good image denoising algorithm, because either fromthe visual effect or objective data, K-SVD algorithm is superior to other similar algorithm.