APPLICATION OF FILTERING EFFICIENCY PREDICTION TO HYPERSPECTRAL DATA PRE-PROCESSING
Several approaches to prediction image denoising efficiency for DCT-based filter have been proposed recently. They allow predicting improvement of PSNR (IPSNR) and visual quality metrics as PSNR-HVS-M (IPHVS) for denoised images under condition of noise characteristics known or pre-estimated in advance. Here we apply the prediction approach to pre-processing ten sub-bands of Hyperion hyperspectral data. It is shown that there are sub-band images for which there is no necessity to carry out filtering. Meanwhile, there are sub-bands for which IPSNR reaches 5…9 dB and the use of denoising is expedient.
Techniques for Earth observation data acquisition, processing and interpretation