Restoration of images with missing pixels
Keníchi Takahashi, Naohiro Ishii · Systems and Computers in Japan · 1991
Abstract At times a portion of the picture elements (pixels) in an image is missing altogether due to a problem with the imaging device or an error in the transmission path, or an empty portion is present due to the fact that a portion of a landscape picture is moved in designing the landscape. In this paper we propose a processing method for restoring the image so that it looks natural in these cases. This scheme consists of the two steps of edge linking in the missing part and subsequent gray level estimation. In the edge linking step, edges which are disconnected in the missing portion are linked, based on an evaluation function representing the validity of edge connectivity. In the gray level estimation step, we interpolate so that there is a pseudo silhouette between each of the missing pixels in each region which is separated by linked edges. Furthermore, in order to obtain a natural feeling in the restored portion, we study the use of an artificial buildup of Gaussian noise. In order to validate the effectiveness of our scheme, we have implemented computer simulations. In the simulations we used a portrait included in SIDBA and made five missing parts, and then tried out the proposed restoration process.