Interpolation based image zooming methods provide a high execution speed and low computational complexity. However, the quality of the zoomed images is unsatisfactory in many cases. The main challenge of super- resolution methods is to create new details to the image. This paper proposes a new algorithm to create new details using a zoom-out-zoom-in strategy. This strategy permits reducing blurring effects by adding the estimated error to the final image. Experimental results for natural images confirm the algorithm's ability to create visually pleasing results.

In this paper a new single image super-resolution is proposed. The proposed method uses an interpolation using an adaptive kernel. The kernel is computed in each position according to the image contents. To evaluate the performance of the proposed method, a comparative experiment with the well-known bi-cubic super-resolution is performed. The experimental results confirm the efficiency of the proposed method.