Multi-Channel Image Denoising In Local Spectral Component Decomposition

Abstract

Based on the low image quality and effect of noise in the conventional methods, a method is implemented for local spectral component decomposition on the line feature of local distribution. The use of local spectral components contributes to achieving better results compared with the result of the stand-alone conventional method. The aim is to reduce noise on multi-channel images by exploiting the linear correlation in the spectral domain of a local region. By calculating a linear feature over the spectral components of an M-channel image, the image is decomposed into three components as a single M-channel image and the two gray scale images. By virtue of the decomposition, the noise is concentrated on the two images and thus the algorithm denoises only the two gray scale images, regardless of the number of channels. As a result, the image deterioration due to the imbalance of the spectral component correlation can be avoided. This method is especially effective for hyper spectral images. Hyperspectral image denoising using a spectral line vector field uses the correlation among spectral information in the local region. The vectors are obtained by the local spectral component decomposition followed by iterative filtering steps. Filtering the spectral line component and the residual component gives significant effects in reducing the noise and smoothing results in the image. The increase in noise power and the number of channels processed affects the complexity of achieving more accurate spectral line vector estimation. This denoising method based on the spectral line is used in remote sensing field. This method improves image quality with less deterioration while preserving vivid contrast.