EVALUATION OF RESTORATION TECHNIQUE IN COMPLEX LANDSCAPE AREAS

In remote sensing images, the problems related to spatial resolution, image degradation and pixel mixture could particularly affect heterogeneous areas. Restoration is a technique that aims to correct radiometric distortions and, combined with a resampling filter, generates images in a finer grid with improved visual quality. This study aims to evaluate the effectiveness of restoration technique to improve quantitative measurements of classification in complex landscape areas. For this purpose, a Landsat 5 Thematic Mapper image with 30 m spatial resolution was processed using restoration and resampling techniques, resulting in a 15 m spatial resolution image. Preliminary results applying a linear spectral mixing model, followed by supervised classification indicated that the restored image showed better visual quality, thus allowing to detect targets in the scene with more details. However, a quantitative comparison between processed and original images, resulted in slight differences (±0.003) in classification accuracy.