Natural Image Reconstruction for Noise-Affected Satellite Images Using ICA

The satellite imagery is one of key source of information gathering associated with environment, natural resource and reconnaissance. As satellite images cover large area and minute details, it is very important to preserve the fine details to obtain highly accurate information. This task is difficult to achieve as images are often affected with noise. So, a preprocessing stage is required which can suppress the level of noise before passing image to applications. This suppression can be achieved by considering each component of data as linear combination of multiple signals. There are many linear transformation techniques available, but ICA is a recently devolved and it is one of the most prominent linear transformations. ICA is a generally utilized for blind source separation technique in which mutual dependence of signal components is minimized. Independent component analysis is applied to numerous applications such as satellite communication, satellite imagery, vibration analysis, speech processing and biomedical signal processing and machinery fault diagnosis. As images are highly influenced by noise due to dual effect of semiconductor aging and image transmission. Hence, it become more critical to remove the noise and reconstruct the natural image by using the image denoising process. The role of ICA in natural image remonstration is expressed in this paper.