The blending interpolation algorithm based on image features

Abstract

In this paper, we develop a blending interpolation model with help of classical bivariate rational interpolation. The blending model is an organic unity which integrates rational fractal interpolation with bivariate rational interpolation, and it is identified uniquely by the values of scaling factor and shape parameters. Fractal is a powerful tool for the description of the complexity and irregularity of geometric objects, and we introduce the fractal dimension to describe texture. A new local adaptive threshold method which based on local fractal dimension (LFD) is proposed to divide images into texture region and non-texture region. Rational fractal interpolation and rational interpolation are used in texture region and non-texture region respectively. Especially in rational fractal interpolation model, a new method for calculating the scaling factor is proposed, which exploits the relationship between global fractal dimension (GFD) and LFD. Finally, an approach of selecting shape parameters is utilized to further improve the quality of interpolated image. Our extensive experimental results demonstrate that the proposed blending model based image features achieves competitive performance with the state-of-the-art interpolation algorithms.