Abstract: In the paper，the authors introduce a method of segmentation of quasi-circular sediment based on Gabor wavelet through texture analyse，in order to segment out the end side of steel bars bundles.The authors define response dispersion for the purpose of choice filter constructed by Gabor wavelet.Furthermore the authors propose Qiao-fuzzy algorithm to optimize the response of the image processed by Gabor wavelet.The authors can unsupervised segment the end side of steel bars bundles form the arbitrary background，by the method of construction，choose of Gabor wavelet and optimization of response.The method is application to a largely image segmentation of quasi-circular Sediment.The method is very stick and very generality.

Key words: texture segmentation, Gabor wavelet, stuff algorithm, quasi-circular sediment

摘要： 以堆积在一起的棒材端面图像为背景，通过纹理分析的方法，提出了一套基于Gabor小波的类圆形堆积物分割方法及其预处理方法。对于Gabor小波构造出来的滤波器组，提出通过响应离散度来优选滤波器；对于得到的频谱响应图像，提出了桥模糊算法来优化频谱响应。通过提出的Gabor小波滤波器构造、选择和优化方法，可以无监督的成功将成捆棒材端面图像从任意背景图像中分割出来。该方法可以应用于该一大类的堆积物分割，具有很强的通用性和鲁棒性。

关键词: 纹理分割, Gabor小波, 模糊算法, 类圆形堆积物