A second-order derivative-based edge preserving image interpolation

Abstract

It is a challenging work to reproduce the high resolution image from the low resolution input while preserving edge structures without causing visual artifacts such as blurring, ringing and jagging. In this paper, we propose an edge guided interpolation method exploiting a data fusion strategy, which interpolates a missing pixel by a weighted sum of directional data estimates for two orthogonal directions. Firstly, each directional data estimate is calculated by 1-D bicubic method along the corresponding direction. Then, we employ the curvature, the second-order directional derivative, to obtain the weights for data fusion. The curvature of the missing pixel decreases as the direction of the curvature is closer to the edge orientation. Experimental results show that the proposed method does not only require the lower computational complexity but also perform better in terms of both subjective and objective visual quality as compared with the state of the art image interpolation algorithms.