Hyperion Image Optimization in Coastal Waters

Abstract

Remote sensing of shallow waters may produce images characterized by limited image coverage, strong uneven background, and high noise/speckle levels, which contribute to the challenges of extracting spatial information. To better assess the submerged aquatic vegetation (SAV) habitat of coastal Pinellas County, Florida, USA, using Hyperion images, two operational image optimization algorithms, vertical radiance correction (VRadCor) for destripe and spectral recognition spatial smooth hyperspectral filter (SRSSHF) for denoise, were modified for use in the shallow coastal waters and then compared to other methods. The VRadCor compresses the cross-track radiance abnormity addressing both the along-track cambering effect with low frequency and the stripe effect with high frequency by estimating both the additive and the multiplicative correction factors. The experimental results show that VRadCor more effectively removes stripes from Hyperion images in comparison to other traditional algorithms. Application of SRSSHF, a special adaptive filter model that compresses the noise by using both spectral and spatial features, was effective for denoising for inner patch areas while retaining (or enhancing) subtle edges between different patches. The use of VRadCor and SRSSHF significantly improves the quality of images of coastal waters while retaining the spectral features of water/SAV. The optimization of the images may lead to improved feature classification or increased accuracy for parameter extraction.