Restoration and segmentation of images by using binding processes

The method of restoration and segmentation of images by using the binding process is proposed. The restored images can keep edges. The binding process uses the relative line process. The binding method can restore multiscale image by the change of a parameter or by the use of diffusion equations. The expansion to impulse noise elimination and application to the restoration and segmentation of sparse data are shown. The advantages of the method over the line processing method are: (1) an invariable result is obtained over a wide range of scale of gray level (2) the segmentation is obtained directly (3) easy elimination of impulse noise, and (4) applicability to the segmentation of sparse data such as dot patterns. © 1998 Scripta Technica. Syst Comp Jpn, 29(4): 79–85, 1998