抄録
Example-based super-resolution is an image interpolation algorithm that uses a database of training images to create plausible high-frequency details in zoomed images. The algorithm is fairly simple; however its performance heavily depends on the database. In particular, when the characteristics of a target image to be magnified are different from the training images, the quality of the super-resolved image degrades. By creating a database consisting of a few training images that closely resemble the target image, we have solved the above problem and improved the performance of example-based super-resolution. This is done by transforming selected images which are downloaded from Internet photo sharing sites, to match their characteristics with those of the target image before adding them to the database. The advantage of this method is that by skillfully creating a database of suitable training images, we are able to improve the quality of the super-resolved image.
