Image continuation

Image continuation, the problem of restoring images that have undefined pixel values at known pixel locations. In document images, undefined pixel values at known locations can have several causes, such as the occurrence of a scratch or blemish on the contact glass of a scanner or the malfunction of a CCD sensor. Once identified, these known locations provide critical information that can be exploited to continue the image from the reliable parts to these unreliable regions. The authors present a family of algorithms accomplishing image continuation, using a small, simple kernel function adaptively applied on successive iterations. Some theoretical properties of this family other than convergence have been established and several examples are presented to illustrate the behavior of the algorithms.<<ETX>>