基于离散余弦变换的水平集算法研究

Level Set Algorithm Based on Discrete Cosine Transform

针对传统水平集算法只对含有加性噪声的图像有较好处理结果的缺点,提出了一种基于离散余弦变换(DCT)的水平集研究算法。首先以图像中的每一点为中心构造隶属于图像的一系列子图像,对这些子图像进行二维DCT变换得到其变换系数,并受Lee滤波器可以去除乘性噪声的启发对这些系数进行运算,得到去除噪声后的平滑图像,再将平滑图像梯度的递减函数作为水平集演化方程的速度停止项,最后根据水平集演化方程对图像进行演化运算。实验结果表明,该方法能够在抑制乘性噪声的同时较好的对图像进行分割。

A new level set algorithm based on discrete cosine transform（DOT） is proposed in order to solve the problem that the traditional level set can only deal with the image with additive noise. For each pixel in the original image,the two dimensional DOT is computed in its local neighborhood. And the denoised image derived from the local DOT coefficients is inspired by the characteristics of the Lee filter. Using the decreasing function of the denoised image gradient as the stop item,the level set function is evolved,and then the final segmentation is produced. Finally,the experimental results show that the developed algorithm could segment the image well while suppressing the multiplicative noise.