Vol. 4, No. 1, 26-43, 2005

Abstract
This paper presents an innovative interpolation method applied to remotely sensed data, based on an appropriate check of the pixels that need interpolation and of the data points used for interpolation. The aim of the method is to project remotely sensed data on a Mercator grid (Datum WGS84) to obtain a highly defined raster image especially along the coastline, where the sensor footprint includes a portion of sea and a portion of land simultaneously.

This interpolation method is mainly based on checking the points acquired by satellite sensor before computing interpolation. This check is based on a criterion of suitability for interpolation depending upon satellite points and regular Mercator grid pixels. For this purpose, we defined some fixed geographical area formats that were accurately segmented into land-pixels and sea-pixels. This result was possible by developing a sub-pixel precise data navigation method for AVHRR (Advanced Very High Resolution Radiometer) NOAA remote sensed data which uses automatic geographical correction. Moreover, considering the dimensions and the shape of AVHRR sensor lobe, for each satellite point the closeness to the coast is taken into account in order to identify mixed points.

Considering land-sea classification, closeness to the coast of satellite points and land-sea classification of Mercator raster image pixels, suitable points for interpolation are selected while unsuitable points are calculated again using close suitable points. The results show a remarkable enhancement in image definition along the coastline.

This method was optimised on a regional scale, in the areas of the Tuscan Sea and Tuscan Archipelago, and its operation was verified using SST (Sea Surface Temperature) images obtained with AVHRR data and also with artificial data. Such testing was carried out through a numerical comparison between images obtained using segmented and ordinary interpolation and also in situ data where present.

ISSN 1729-3782