Noise Filtering in High-Resolution Satellite Images Using Composite Multiresolution Transforms

This contribution proposes a multiresolution analysis (MRA)-based composite technique for image restoration by noise filtering in satellite images. Multiresolution techniques provide a coarse–fine and scale-invariant decomposition of images for analysis and interpretation. MRA methods effectively handle the noise because of their multiscale feature. This study presents a scheme based on the combination of wavelet-, contourlet- and curvelet-based transforms as effective tool for noise filtering in satellite images. The proposed method is applied to the problem of restoring an image from noisy data and effects of denoising are compared. Several comparison experiments with state-of-the-art noise filtering schemes are conducted. The composite approach of curvelet and wavelet is found to be more effective than the others based on the set of evaluation measures like peak signal–noise ratio, mean-squared error, edge-enhancing index and mean–standard deviation ratio across edges. The results are illustrated using high-resolution satellite data, such as Quickbird and Worldview-2 images. Such high-resolution images are more likely to be noisy due to the short observation time over the target in contrast to images from low-resolution sensors.