PROJECT TITLE :

Single Image Super-Resolution Based on Gradient Profile Sharpness - 2015

ABSTRACT:

Single image superresolution could be a classic and active image processing problem, that aims to get a high-resolution (HR) image from an occasional-resolution input image. Thanks to the severely underneath-determined nature of this problem, a good image prior is critical to create the matter solvable, and to improve the standard of generated images. During this paper, a novel image superresolution algorithm is proposed based on gradient profile sharpness (GPS). GPS is a position sharpness metric, which is extracted from two gradient description models, i.e., a triangle model and a Gaussian mixture model for the outline of different sorts of gradient profiles. Then, the transformation relationship of GPSs in different image resolutions is studied statistically, and the parameter of the link is estimated automatically. Based mostly on the estimated GPS transformation relationship, two gradient profile transformation models are proposed for two profile description models, which can keep profile form and profile gradient magnitude sum consistent during profile transformation. Finally, the target gradient field of HR image is generated from the remodeled gradient profiles, that is added as the image prior in HR image reconstruction model. Extensive experiments are conducted to guage the proposed algorithm in subjective visual impact, objective quality, and computation time. The experimental results demonstrate that the proposed approach will generate superior HR pictures with higher visual quality, lower reconstruction error, and acceptable computation potency as compared with state-of-the-art works.