改进的数字图像轮廓曲线傅里叶描述算法

An Improved Fourier Descriptor Algorithm for Digital Image Contour Curve

研究图像轮廓曲线特征问题,图像轮廓曲线的描述是图像理解的一个核心问题,傅里叶描述算法是图像轮廓曲线描述的较优算法。由于传统的轮廓曲线傅里叶描述算法中曲线的采样过于复杂和不稳定,降低了识别的性能。为提高曲线描述的性能,提出了一种改进的图像轮廓曲线傅里叶描述算法,通过运用循环移位采样算法对曲线采样得到多组采样点,并取傅里叶变换后最小的一组规范化系数作为曲线的傅里叶描述子,避免了传统算法中复杂的采样过程,使算法更简洁稳定。通过算法仿真与实验对比,算法有更高的准确度和稳定性,能有效反映轮廓曲线特征。

Description of two-dimensional image contours is a core issue of image understanding.Fourier curve description algorithm is the optimum algorithm to describe curves.The curve sampling in the traditional Fourier curve description is too complex and unstable,which reduces the performance of the algorithm.To improve the performance of the curve descriptor,an improved algorithm of Fourier contour curve description was proposed in this paper.Through using the cycle shift sampling algorithm to obtain multiple sets of curve sample points and taking the minimum set of standardized coefficients after Fourier transform as the Fourier descriptor of the curve,our algorithm avoided the complex sampling process of the traditional algorithm,and made the algorithm more concise and stable.Through the algorithm simulation,compared with the traditional Fourier curve description algorithm,experimental results show that this algorithm can describe the curve more accurately,reflect the differences between the different curves effectively.