Compression based refinement of spatial interpolation of missing pixels in grey scale images

We present a method for refining interpolated values for pixels in grey scale images that were corrupted during transmission. The method uses an adaptive image model developed in the context of image compression, allowing it to exploit prior knowledge of image characteristics that can either be derived from examples coming from the same class of images or alternatively be determined by the sender and transmitted separately. The proposed method has been tested in a number of experiments and has been shown to improve both objective quality measures as well as subjective quality over a simpler non-adaptive method