Super Resolution from a Single Image Based on Self Similarity

A single image super resolution method based on self-similarity is proposed. It has been proposed that self-similarity occurs in the same image scales as the source image as well as across different image scales. We exploit this novel innovation to generate the image database, which is also referred to as image codebook, from two scales of the source image representing the relationship between low-frequency and high-frequency components. Thus, the missing high-frequency details in the resultant image to be enlarged are estimated by finding the nine nearest neighbors in the database with the low-frequency coming from the interpolation method. Then, the nine corresponding high-frequency components are weighted to add to the low-frequency to obtain the super-resolution image. Finally, a back-project operation is applied on the resultant image to ensure that the enlarged image is consistent with the source image as well as possible.