Superresolution border segmentation and measurement in remote sensing images

Segmentation and measurement of linear characteristics in remote sensing imagery are among the first stages in several geomorphologic studies, including the length estimation of geographic features such as perimeters, coastal lines, and borders. However, unlike area measurement algorithms, widely used methods for perimeter estimation in digital images have high systematic errors. No precision improvement can be achieved with finer spatial resolution images because of the inherent geometrical inaccuracies they commit. In this work, a superresolution border segmentation and measurement algorithm is presented. The method is based on minimum distance segmentation over the initial image, followed by contour tracking using a superresolution enhancement of the marching squares algorithm. Thorough testing with synthetic and validated field images shows that this algorithm outperforms traditional border measuring methods, regardless of the image resolution or the orientation, size, and shape of the object to be analyzed.

► A superresolution border segmentation and measurement algorithm is introduced. ► Traditional methods produce an underlying vectorization with systematic errors. ► They are used in most commercial and open source geographic information systems. ► Some segmentation results given by the Envi did not allow for any measurement. ► Our method is more accurate and less sensitive to changes in shape and orientation.