Application of Merging Pixels in Restoring Mixed Pixels Data

Mixed pixels are one of the main problems in the remote sensing data classification. There are various reasons including limitations of sensor spatial resolution or costs of data acquisition that may cause the image accuracy to be less than desired quality. A similar problem may be in comparing different spatial resolution images of one scene. Increasing the spatial resolution in such cases is necessary. Different methods have been proposed to increase the spatial resolution with maximum accuracy. These studies considered proportion of subpixels and locating their positions. These methods and their problems will be examined in this study. Conventional methods are reviewed. And a new method for increasing the spatial resolution will be proposed to resolve some of the weaknesses of existing methods. This new algorithm is proposed for restoring subpixels in the hyper spectral images. This algorithm is based on mixture of pixels for creating a proper search table for restoring subpixels. The spatial and pattern properties of neighboring pixels are considered in this search. There is no need to use soft classification information. Testing algorithm on real data will show its performance and capability in remote sensing applications. Simulation results on real data show that proposed method can increase percentage of correction classification at least 15% relative to hard classification.