基于分形维数的复杂图像特征提取方法

EXTRACTION OF COMPLEX IMAGE FEATURES BASED ON BOX DIMENSION

提出了一种通用的图像特征提取方法 ,这种方法运用分形维数可以表示曲线的复杂程度的特性来提取图像纹理特征 ,它具有计算简单、物理意义明确、抗干扰能力强的特点 .证明了对于离散函数图像在一定条件下可以采用与连续函数图像类似的方法求分形维数 .最后给出了一个应用实例表明不同复杂程度的纹理其频谱曲线的分形维数是不相同的 .

Describes a novel method to extract image features,which uses fractal dimension to indicate the complexity of a plot,and features simple calculation,definite physical meaning and robustness,and proves that under some hypothesis the method of continuous plot can be used to seek the dimension of a discrete plot,and gives an example to indicate that the fractal dimension of plot of S (θ) differs from the complexity.