Representation and Denoising Based on Non-negative Sparse Coding

This paper proposes a new non-negative sparse coding (NNSC) model. And using this model, the features of natural image are extracted successfully. As a practical application, how to reduce noise added in a natural image is also discussed. Our assumptions are that the image is corrupted by additive and multiplicative white Gaussian noise and uniform distribution random noise. According to our experimental results, it is clear that NNSC algorithm performs well in terms of image visual effect and the SNR values. As well as, NNSC technique compares very optimally against some of the most recently published denoising such as wiener filter, wavelet threshold shrinkage, standard linear sparse coding (SC) etc.