Primal sketch of image series with edge preserving filtering application to change detection

This paper proposes an original approach to the problem of change detection into multitemporal image series. A primal sketch of the image is achieved with a modified Perona-Malik filter: relevant edges are extracted. Then, an application to change detection is proposed. The algorithm related to sketch outputs is performed through density of pixels images. It offers significant information about events that occurred between several acquisitions. Potential of the algorithm on high resolution images is shown, and limitations on very high resolution are explained.