Remote Sensing

Remote sensing is a technology for gathering data and information about the physical world by detecting and measuring signals emanating from specific source targets on the Earth's land surface, oceans, and atmosphere. Since the final product of remote sensing is usually an image displaying some physical properties of the observed area, image processing techniques play a very important role in the production, organization and interpretation of data acquired by remote sensing technologies.

The experience of the IAPP group in image processing for remote sensing is mainly related to image analysis, enhancement and restoration techniques for data coming from synthetic aperture radar (SAR) and optical/infrared hyperspectral sensors.

- SAR Image Processing

SAR remote sensing offers a number of advantages over optical remote sensing, mainly the all-day, all weather acquisition capability. However, the main drawback of SAR data is the presence of speckle, a signal dependent granular noise, inherent of all active coherent systems, which visually degrades the appearance of images and may severely diminishes the performances of automated scene analysis and information extraction techniques.

The availability of efficient tools for speckle reduction, or de-speckling, may become of crucial importance in applications requiring multiple SAR observations. Since the pioneering Lee's filter in 1980, an extensive literature has flourished around the problem of de-speckling. The last ten years have witnessed the ever increasing presence of de-speckling methods based on multi-resolution analysis. The superior performances of such methods over conventional spatial filters has been recognised only very recently for de-speckling, unlike for standard image de-noising, which has started benefiting from multi-resolution approaches several years before.

- Hyperspectral Image Processing

Hyperspectral sensors collect image data simultaneously in dozens or hundreds of narrow, adjacent spectral bands. These measurements make it possible to derive a continuous spectrum for each image cell, which can be compared with field or laboratory reflectance spectra in order to recognize and map a large variety of surface materials/materials. Due to their potentials, hyperspectral images can find many applications in resource management, agriculture, mineral exploration, and environmental monitoring.

Recently, the ability to simultaneously capture a high-resolution panchromatic (PAN) image and a hyperspectral image, typically with much lower resolution, has been taken into consideration. All this makes the development and evaluation of algorithms that allow to achieve better spatial resolution in hyperspectral images through image fusion techniques very interesting.

For the proper functioning of these techniques it is extremely important to characterize the sensor noise model, which in modern sensors often involves the prevalence of photonic noise, which is signal dependent and intrinsic of the optoelectronic acquisition process, compared to electronic noise, which is independent of the signal and is likely to be reduced with the progress of technology. In addition, techniques of image enhancement and restoration based on these noise models can be effectively used to improve the quality of data.