Remote Sensing Image Segmentation and Representation through Multiscale Analysis

Every year, new sensor technologies are being implemented to improve the acquisition of high-resolution remote sensing images (RSIs). With the large amount of data provided by these sensors, novel computational approaches are constantly required to support the decision-making process based on RSI analysis. A typical problem is the recognition of target regions for land-cover mapping. In this context, the main problems are: (1) classification methods are dependent on the segmentation quality; and (2) the selection of representative samples for training is a costly process. The samples indicated by the user are not always enough to define the best segmentation scale. Furthermore, the indication of samples can be expensive, since it often requires to visit studied places in loco. The segmentation-dependence problem has been addressed in the literature by using multiscale analysis. The training sample selection problem is, in turn, addressed mainly by employing user interaction techniques which are usually combined with pixel-based classification approaches. This work aims to introduce problems, challenges, and some state-of-the-art approaches for multiscale classification of remote sensing image. The main covered topics are arranged into four sessions: research challenges, segmentation, feature extraction, and classification.