基于Gabor变换的纹理图像分割算法及应用

Research on Texture Image Segmentation Algorithm and Its Application Based on Gabor Wavelet Transform

基于Gabor小波的多尺度、多方向性,结合模极值特征提取方法和聚类算法提出一种新的纹理图像分割算法,并将算法应用到竹材横端面的识别中。二维Gabor小波可以在空域、频域和方向上获得最佳的分辨率,利用其对纹理图像进行分解,提取图像模极值纹理特征,用聚类算法对纹理图像进行分割。实验结果表明,提出的方法对合成纹理图像有理想的分割效果,并能够很好地对自然纹理图像进行分割。算法在竹材横端面的分割识别中,取得了很好的效果。

Based on multi-scale and multi-direction of Gabor Wavelet Transform,a new texture image segmentation algorithm was presented with modulus extremum feature extraction and K-means clustering algorithm in this paper.The algorithm was applied to the recognition of the bamboo cross-section.The 2D Gabor wavelet can get the best resolution in the spatial domain,frequency domain and the direction.Gabor wavelet was used to decompose texture image and extract the texture features by modulus extremum.At last,texture image was segmented by K-means clustering algorithm.Experiment result showsed that the method can be good for synthetic and natural texture images segmentation.The algorithm has achieved good results in the segmentation and recognition of the bamboo cross-section.