摘要： Hao-Cecilia图像插值算法能消除边缘锯齿，但只能进行2倍插值。针对该问题，提出一种改进的线性图像插值方法。对图像插值变分正则化约束的欧拉方程用泰勒展开式进行离散化，使插值沿着图像水平线方向进行，消除插值图像边缘的锯齿现象并抑制模糊。实验结果显示，该算法能产生良好的视觉效果，使客观评价指标(峰值信噪比)得到较大提高。

关键词: 图像插值, 图像缩放, 泰勒展开式

Abstract: The Hao-Cecilia image interpolation algorithm can eliminate the zigzagging artifacts better, but the shortcoming of the expansion ratio of 2 is inherent. In this paper, a novel linear image interpolation method is proposed to overcome it. The Euler equation of the variational regularizer is discretized using a Taylor expansion. The interpolation takes place along the direction of level line, which eliminates the zigzagging artifacts and suppresses blurring of edges. Numerical experiments demonstrate that the proposed algorithm can obtain better visual effect and higher Peak Signal Noise Ratio(PSNR).

Key words: image interpolation, image zooming, Taylor expansion

中图分类号: 

• TP393