基于小波变换与图割的彩色图像分割方法

Color Image Segmentation Based on Wavelet Transform and Graph Cuts

为提高图割算法的分割效率与质量并改善shrinking bias现象,提出将图割理论与小波变换相结合的方法.该方法利用小波变换多分辨率分析的特点,将变换中的低频子带图像作为估计GMM参数的训练样本进行多尺度迭代分割,提高算法效率,利用简单高效的CS_LBP纹理描述子提取高频子带图像中的纹理信息,将颜色与纹理特征相结合改善分割效果,并利用高频系数进行多尺度边缘检测,用于计算局部自适应的正则化参数,改善对细长边界的分割.实验结果表明,分割效果得到了改善,算法效率得到了提高.

To improve the efficiency, effect and shrinking bias phenomena of graph cuts algorithm, an approach of combing graph cuts algorithm and wavelet transform is proposed in this paper. By using the character of wavelet multi-resolution analysis, the low-frequency sub-band images of wavelet transform are used as training samples to estimate GMM parameters with multi-scale iterative segmentation efficiently, texture features which are extracted from high frequency sub-band images by using simple and efficient CSLBP texture descriptor are combined with color features to improve the segmentation effect and high-frequency coefficients are used to detect multiscale edges. The local adaptive regularization parameter is calculated with the edge probability map to improve the thin boundary image segmentation. The experiments show that segmentation result has been improved and the efficiency of the algorithm has been improved significantly.