Motion video KNN-SVD denoising algorithm

A top quality and reliable algorithm is proposed for video denoising corrupted by additive zero mean gaussian noise with a well known and fixed variance. The planned methodology consists of three basic steps. Firstly, the strategy classifies similar frame patches by the K nearest neighbour (KNN) technique so as to create similar patch groups. K-Nearest Neighbour (KNN) clustering algorithm considerably reduces the complexity of using other similarity metrics. Secondly, all the clusters in this cluster are denoised by Singular value Decomposition transform-based approach and low rank approximation that separates noise and signal coefficients. Lastly, an initial denoised frame is made by adaptive weight aggregation technique and back projection method. The results demonstrate the flexibility of this methodology to preserve fine textures. Experimental results show that the system is comparable the other strategies in removing AWGN, and considerably outperforms existing techniques in removing real, structured noise.