梯度小波纹理模型及其在超声心动图分割中的应用

The Gradient Wavelet Texture Model and Its Application in Echocardiogram Segmentation

本研究在图像的高斯马尔可夫随机场 (GMRF)模型和梯度小波变换的基础上提出了梯度小波多尺度纹理分析的概念 ,推导出了梯度小波变换系数和高斯马尔可夫过程参数的关系 ,并在此基础上给出了纹理特征参数的估计方法。这些参数构成了一个多尺度纹理特征空间 ,本研究利用K 均值算法实现了特征参数聚类。本研究提出的方法在超声心动图的分割中取得了令人满意的效果。

This paper presented a multiscale texture analysis concept based on the Gauss Markov random field (GMRF) model and gradient wavelet transform. The relation between the coefficients of gradient wavelet transform and the parameters of GMRF was deduced. Based on the above result, the parameters were estimated. These parameters formed a multiscale texture feature space, and the K-means algorithm was applied in the clustering of the parameters. The proposed method showed a satisfied result in the segmentation of the echocardiograms.