运用可变模板进行并行图像处理的一种快速算法

提出了一种在并行机上有效地计算（空间）可变模板的方法，论证了利用一个在图像网格点处计算多项式的优化算法，可以大大减少可变模板的运算量，对于包含非多项式函数的可变模板，可以用函数的泰勒级数展开实现在像素点上的递推运算，详细分析了可变模板中若干常用函数的泰勒展开用于实现模板运算的合理性，准确性和有效性，关于硬件的影响以及该方法的适用范围，也做了讨论。

An Efficient Algorithm for Parallel Image Processing with Variant Templates

A fast algorithm for parallel image processing with variant templates is presented in this paper. This algorithm is based on a key observation that we have a priori knowledge of the changing patterns of the arguments i, j and k, and can take full advantage of this. It is shown that the point wise updating of the variant templates can be greatly simplified by use of an optimized algorithm for evaluating polynomials at the image grid points. While the efficient computation of polynomials plays a central role in this approach, the technique is extended to include transcendental functions with the help of the Taylor expansion. Examples of typical transcendental functions show that reasonably high precision, for common low level image processing applications, and efficiency, using only very few initial terms of the Taylor series, can be achieved, when the series form of the function is used for executing the variant template operations. The influence of hardware and the limitations of the proposed algorithm are discussed at the end of the paper.