摘要
以马尔可夫随机过程和梯度小波变换为基础提出了梯度小波纹理模型。由梯度小波纹理模型的参数组成描述图像纹理的特征向量,梯度小波纹理模型利用了马尔可夫随机场图像模型的优点和基于该模型的成熟方法,并且引入了多尺度、多分辨率等特性。然后采用Kohonen自组织SOM(Self-Organizing Feature Map)神经网络对纹理特征向量进行无监督学习,最后对超声心动图像进行纹理分割,取得较满意的分割效果。
A gradient wavelet texture model is presented based on Gauss Markov random field and gradient wavelet transform. The parameters of the model form a multiscale texture feature space, and provid a multiscale description of image texture. This method takes the advantages of Markov random field image model and imports the multiscale feature of wavelet transform. Kohonen SOM neural network algorithm is applied in the classification of the texture feature vector, thus we segmented echocardiograms and got satisfied result.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2005年第4期474-479,共6页
Pattern Recognition and Artificial Intelligence
