Denoising Method On Low Rank Matrix Completion And Matrix Recovery Occluded Images

In many digital image applications, low-rank matrix sparse representation is wide range of applications at present, such as character recognition, video image editing, face recognition system, 3D scene modeling. Although a large number of classic algorithm can be applied to under the factors such as geometric transformation and noise interference, due to not correctly or close to describe images, it makes solution instability. This paper proposes a new model combined low-rank matrix recovery with low-rank matrix completion.Some algorithms have been proposed to solve matrix completion with matrix recovery problems. Such as Singular Value Threshold(SVT) algorithm, APG algorithm, Iterative Threshold(IT) algorithm, but the algorithm convergence speed and the problem of singular value decomposition, the amount of calculation is very large. Lots of experiments show that convergence speed of Inexactly Augmented Lagrange Multipliers(IALM) more rapidly and higher precision than other algorithms. On selecting the image, the image must be low rank, namely the data distribution is on a lower dimensional linear subspace. Low rank images may exist deletion and noise pollution, simultaneously it gets rid of the noise to obtain the full image using low-rank matrix recovery. In this paper the new model conducts the convex relaxation, then using Inexactly Augmented Lagrange Multipliers(IALM) method to solve the problem. At last, we can get the full image.Analysis the three different images with noise, compared recovery time, signal-to-noise ratio, peak signal to noise ratio and error rate with the existing low-rank matrix algorithm, it can be seen that the improved low-rank matrix completion with recovery model can get rid of the cover noise images very well. In the repair of obscured images, change the model and recover the cover images in this paper. Add up to low-rank matrix completion which is based on low-rank matrix recovery, it makes remove the cover, at the same time, fill the missing parts. This can achieve the ideal effect.