一种图像纹理分割的小波—像素曲线演化方法

A Wavelet Curve Evolution Method for Image Texture Segmentation

提出了图像纹理分割的小波—曲线演化方法,该方法运用二维小波对纹理图像进行分解,由四个小波系数描述纹理特征,并组成四维小波系数特征矢量图,然后采用最大后验概率模型的曲线演化方法,对特征矢量图进行分割和分类,从而获得原图像纹理分割结果.实验结果表明,与小波—FCM聚类纹理分割方法相比,该方法对双纹理图像能取得较好的分割效果,分割出的边界连续,并且具有较强的抗噪能力.

An approach to image texture segmentation is proposed based on wavelet and curve evolution. 2D wavelet decomposes the texture image into four wavelet coefficients which indicate the texture feature and form 4D vector image. In order to get the segmentation result of the original image texture, the wavelet coefficient vector image is segmented and classified by using curve evolution based on the maximum posterior probability(MAP)model. Compared with wavelet FCM method, the experiments have proved that the proposed method can get better results such as that the segmented edge is continuous, and it is robust to noise.