摘要
为了最大程度消除图像放大后出现的锯齿现象,提出了一种新的二次导数连续的多结点立方卷积插值函数,通过增加结点使得插值曲线更加的平滑;同时对插值函数引入逆梯度约束条件,然后根据图像放大的倍数定义一个适当的模板对图像进行卷积操作,自适应的消除图像放大后出现的锯齿现象。实验结果表明,该算法在对图像执行高倍放大时,几乎看不到锯齿,同时图像的清晰度也保持的较好。
In order to eliminate the sawtooth gained by the image magnification, a new second derivative continuous cubic convolution interpolation function of multi-knot is proposed. According to adding knot to make the interpolation curve more smooth, a constraint condition of inverse gradient is introduced. Then in order to eliminate the sawtooth self-adapting in magnified image, a proper template is defined to perform convolution operation based on the enlargement factor. The experiment results show that it is hard to see the sawtooth in the process of high magnification of image, and the definition of image is preserved well.
出处
《四川理工学院学报（自然科学版）》
CAS
2013年第3期34-37,共4页
Journal of Sichuan University of Science & Engineering(Natural Science Edition)
基金
人工智能四川省重点实验室基金项目(2011RY008)
关键词
立方卷积
逆梯度
高倍放大
清晰度
cubic convolution
inverse gradient
high magnification
definition
