Hybrid DCT-Zernike moments-based approach for image up-sampling

Abstract

In this paper, we propose an up-sampling technique in which DCT and Zernike moments (ZMs) are used in hybrid mode to up-sample the images that are down-sampled in DCT domain. DCT-based up-sampling techniques are very effective and simple for up-sampling in DCT domain. However, being discrete and finite, they do not offer any direct way to recover missing details (for instance, substitution of high frequency DCT coefficients by zero values in DCT-based zero padding approach). ZMs, on the other hand, are continuous moments having infinite number of orthogonal bases that can effectively preserve the fine details of an image while reconstructing it. Therefore, the main aim behind this hybrid approach is to combine the strengths of DCT and ZMs to obtain a high quality up-sampled image with minimum ringing and blocking artifacts. Experiments performed on ten different natural images prove the superiority of the proposed hybrid approach over the existing DCT-based up-sampling approaches.