一种改进的C-V水平集遥感图像分割方法

Improved C-V Level Set Segmentation Method for Remote Sensing Images

针对遥感图像边界模糊分割难的问题,提出了一种改进的基于小波变换的C-V水平集分割方法提高其分割准确性。该方法首先使用小波变换得到原图像的高频分量,初步定位图像中高频信号的空间位置;然后根据高频分量的幅值及其空间分布,借鉴反锐化掩模法的思想,增强图像高频信号对水平集分割的指导作用,并优化驱动水平集演化的内、外能量及曲线长度约束能量。实验证明,运用该方法对遥感图像的分割结果比传统水平集方法更准确,能有效的利用局部信息提高水平集能量项的有效性和分割结果的准确性。

A method of improved C - V level set segmentation based on wavelet transformation for remote sensing images with obscure edges is proposed. First, high frequency component of an image is extracted by wavelet decomposition, by which the position of the high frequency signals of the image is located. With the idea of unsharp masking method, the energies （inner, outer and arc length） of C -V level set are recalculated according to the high frequency component and its space-frequency relationship. Experiments show that the segmentation results by this method are more accurate than the traditional ones. So the proposed one can use more reasonable local information for segmentation than the traditional model, and improve the effectiveness of level set energies and the correctness of segmentation results.