Directional denoising of aerial images by splitting-signals

Aerial imagery is a very important data source used in underwater oceanographic studies. These images are usually corrupted by clutters caused by surface water waves. Removal of the wave clutters from these images is an important preprocessing step for accurate assessment of information since the clutters interfere with useful information over the surface. In this work we propose a novel method combining the tensor transform by wavelet transform and denoising these images. Surface waves are classified in two categories: ripple waves and spark waves. Both types of clutters are directional and can be detected by the tensor representation of the image. To remove the long-waves, the subband-filtering technique is used, but it is reduced to 1-D filtering by means of the splitting-signals in the tensor representation. To remove the short-waves, the well-known SMEME (spectral spatial maximum exclusive mean) filter is used with the tensor transform as a pre-step of the filter.