A Novel Sensing Noise and Gaussian Noise Removal Methods via Sparse Representation Using Svd and Compressive Sensing Methods

Image processing is one of the common research areas in recent decades, since noisy images cause harmful consequence on several applications and considerably degrade visual quality. The term denoising indicates to the method of estimating the unidentified (original) signal from available noisy data. Hyperspectral imaging has been established that it has several applications in farming, diagnostic medicine, and military surveillance. On the other hand, in these applications, the occurrence of the noise considerably reduces the classification accuracy. In order to solve these setbacks, for HSI, there is much global and local redundancy and correlation (RAC) in spatial/spectral dimensions is proposed in earlier work to eliminate noise from samples. Additionally, denoising performance can be enhanced significantly if RAC is exploited professionally in the denoising process. Nevertheless, the available RAC method denoising performance possibly will decrease when noise is strong. It turns out to be one of the important issues on dictionary learning. With the intention of surpassing these setbacks, in this paper presented a noise removal scheme to eradicate noise from image samples, at first the sensing noise in the image samples are eradicated with the help of the Singular Value Decomposition (SVD) and the Gaussian noise in the image samples are eradicated with the help of the Compression Sensing (CS) methods. SVD algorithm utilizes both the spectral and the spatial information in the images. Noise can be eliminated by sparse approximated data with SVD techniques. The denoising outcome from the proposed method is better than the other hyperspectral denoising schemes. Our results demonstrate that our denoising method can achieve competitive performance than other state-of-the-art methods.