关键词: 图像放大, 多项式逼近, 自适应区域, 函数插值

Abstract: In order to improve the sawtooth gained by the traditional image zooming algorithm, this paper analyzes the statistical characteristics of the image pixels, and divides the image into smooth, texture and edge. Based on the local homogeneity in smoothing and texture area, the paper proposes image zooming algorithm based on local polynomial approximation. Firstly, it divides the image into several homomorphic image regions using the intersection of the confidence interval algorithm. Secondly, it figures the best polynomial fitting for each area, and then making oversampling of the fitting function to achieve regional amplification. Finally, it uses local smoothing to obliterate jagged edges. The experimental results show that the algorithm makes full use of the structural information of the image and improves the image visual effects and applicability compared with conventional interpolation zooming algorithm and the existing locally adaptive interpolation algorithm.

Key words: image zooming, polynomial approximation, adaptive area, function interpolation