A Survey of Multispectral Image Denoising Methods for Remote Sensing Imagery Applications
In contrast with the conventional RGB or gray scale images, the multispectral images tends to convey more faithful representation for real world scenes to enhance the performance of many tasks entail with computer vision, object extraction, detection and quantification, tagging operations and image segmentation. High fealty of color reproduction is possible by using multispectral images of visible spectrum than the normal RGB systems due to attainment of limited information conveyed by RGB images. While capturing, the MSIs are certainly corrupted by various noises which may be due to limitations in equipment, scanty bandwidth and loss of radiant energy. Formulating a novel mathematical description of deep learning based denoising model is a complex research question and many researchers specified different algorithms or methods for denoising of MSI. Many researchers have suggested its use with the application of neural network as a sparse coding of noisy patches. Moreover, these allow various algorithms to amend itself for a task using machine learning algorithm. However, in general practice, a multispectral image is always encountered by various noises. In this study, we presented the past techniques specified for the noise influenced MSI. The survey describes the overview of past techniques and their advantages in comparison with each other.