Abstract: The current gradient algorithm only consider the gradient between the two color images without considering the inherent correlation between the three color components of the problem. To solve this problem, a new nonlinear method of image metamorphosis among multiple images based on bivariate rational interpolation was represented. First, a Newton-Thiele’s vector valued interpolation surface was constructed for the values of RGB pixels, and then series of morphing images were generated by resampling the surface of this interpolation function. The experimental results show that this algorithm is better than several other algorithms both in integrality of image feature and visibility of transition images.

Key words: color image metamorphosis, non-linear interpolation, rational interpolation, vector valued continued fraction, blending interpolation

摘要： 针对目前图像渐变算法只考虑两个彩色图像之间的渐变并且没有考虑三个颜色分量内在相关性的问题,在二元混合向量有理插值的基础上，提出了一种非线性的多幅彩色图像渐变的新方法。首先将多幅图像中每个像素的RGB三原色看做是平面域上的向量，利用二元Newton-Thiele型向量连分式建立有理插值曲面，再对此插值曲面进行重采样，得到一系列的渐变中间图像。实验结果表明,该算法在保持图像特征和过渡图像的可视性方面均优于其他算法。

关键词: 彩色图像渐变, 非线性插值, 有理插值, 向量连分式, 混合插值