LEARNING TRADITIONAL FILTERS BASED ON IMAGE EXAMPLES

Abstract

This paper presents an approach for image filtering process depends on image examples. The approach involves of preparing a training images, in which a pair of images, with one image purported to be a filtered version of the other, as a training data. The filtered property is to be automatically learned and transferred to another target image. The basic idea behind learning approach is depended on luminance neighborhood statistics. Statistics pertaining to each pixel in the target pair are to be compared against statistics for every pixel in the source pair, and the closest match is to be found and its property (i.e., luminance information) is applied to the target image in order to create a filtered image. This approach is simple and provides results comparable to that obtained in image analogies of the Hertzmann et al.