Image interpolation by high dimensional projection based on subspace method

This work presents a new image interpolation technique for image cracking such as telop character occlusion or picture deterioration. The landscape picture and texture pattern or other natural scenes contain the behavior of the autocorrelation pattern, and a method of image interpolation based on the autocorrelation pattern has been established. This method extracts many local regions from a piece of image and generates a subspace by the image vectors. The subspace limits the freedom of the image description at the local region by the hyperplane restriction in the image vector space. This paper explains for an image interpolation method based on eigen-space and explains the expansion using kernel non-linear projection.