The Estimation Of Image Noise

This thesis mainly estimates the Variance of images with Gaussian White Noise in space domain.Drawing lessons from the principals of the similar blocks matching and noise estimation and filtering using block-based singular value decomposition, we adop-t the Linematch-method to estimate the image with noise. And the algorithm’s main ideas are as follows:firstly, we add Gaussian White Noise to images without noise,then look for similar lines in the images with Gaussian White Noise,and make group with the similar line into similar line matrix,then decomposing the above similar line matrix on the singular value decomposition; furthermore introducing the BP-method to learn the singular value; finally, we look for similar lines in the unknown Gaussian White Noise image,and make group with the similar line into similar matrix,then decomposing the above similar line matrix on the singular val-ue decomposition,make use of the above learning result to simulate to obtain the estimation of image noise.There is mainly some difficulty during the noise estimation algorithm:to make sure the stability of the algorithm.Besides,how to avoid the image information in-fluencing on the noise; In addition, the generality of the algorithm is also hard to solve.