Analogic CNN algorithms for some image compression and restoration tasks

摘要

This paper presents analogic cellular neural network universal machine algorithms for some important image processing tasks. Templates are given for skeletonizing black-and-white and grey-scale images. A method is presented enhancing the quality of copying machines by removing scratches from the copied images. Finally, a similar method is used to enlarge images. All templates use only a nearest neighborhood, therefore being well suited for VLSI implementation.