Web 2.0, personal learning environments and the future of learning management systems

Many applications in image analysis are concerned with the temporal evolution of shapes in video sequences. In situations involving low-contrast, low-quality images, human aid is often needed to extract shapes from images. An interesting approach is to use expert help to extract shapes in certain well-separated frames, and to use automated methods to extract shapes from intermediate frames. We present a technique to interpolate between expert generated shapes. This technique preserves salient features in the interpolated shapes, and allows analysts to model a continuous evolution of shapes, instead of a coarse sampling generated by the expert. The basic idea is to establish a correspondence between points on the two end shapes, and to construct a geodesic flow on a shape space maintaining that correspondence. This technique is demonstrated using echocardiagraphic images and infrared human gait sequences.