论文部分内容阅读
本文提出了一种用Systolic结构实现图像几何校正的方法,我们采用多项式来逼近几何校正的变换函数,由于坐标变换在图像平面上进行,多项式的计算可以大幅度简化。我们给出一个叠代算法,当用systolic结构来实现这个叠代算法时,计算一个象素点的坐标只需一次加法的时间,完成全部计算的加法器数目与多项式的阶次成线性关系。对运算的精度,我们作了理论上的分析。根据这个方法,我们用硬件实现了二阶多项式函数几何校正,几何校正的计算可以以接近实时的速度运行。
In this paper, we propose a method of image geometric correction using Systolic structure. We use polynomials to approximate the transformation function of geometric correction. Since the coordinate transformation is performed on the image plane, the polynomial calculation can be greatly simplified. We give an iterative algorithm. When the systolic structure is used to implement this iterative algorithm, the coordinates of a pixel are calculated in one single addition time. The number of adders to perform all calculations is linear with the order of the polynomial. The accuracy of the operation, we made a theoretical analysis. According to this method, we implement the geometric correction of second-order polynomial functions in hardware, and the computation of geometric correction can be run at near real-time speed.