文摘
Remote Sensing Images are one of the main sources of information about the earth surface. They are widely used to automatically generate thematic maps that show the land cover of an area. This process is traditionally done by using supervised classifiers which learn patterns extracted from the image pixels annotated by the user and then assign a label to the remaining pixels. However, due to the increasing spatial resolution of the images resulting from advances in the acquisition technology, pixelwise classification is not suitable anymore, even when combined with context. Therefore, we propose a new descriptor for superpixels called Star descriptor that creates a representation based on both its own visual cues and context. Unlike the most methods in the literature, the new approach does not require any prior classification to aggregate context. Experiments carried out on urban images showed the effectiveness of the Star descriptor to generate land cover thematic maps.