基于概率纹理模型的纹理分割方法分析

ANALYSIS OF PROBABILISTIC TEXTURE MODEL BASED TEXTURE SEGMENTATION METHOD

基于最优小波包基的纹理自适应概率模型通过优选小波包基来区分不同纹理,具有纹理描述更准确的优点。研究基于纹理自适应描述的邻域分割法,通过实验分析了纹理概率模型和邻域分割法对分割效果影响的主次关系。实验结果表明邻域分割法是分割取得好效果(分割错误率低于1.34)的主要影响因素,概率模型对分割效果的作用是次要因素。这一结论的得出将有利于对该方法的改进。

The probabilistic adaptive texture model based on best wavelet packet basis has excellence in more accurate texture description which differentiates various textures by optimizing wavelet packet basis. An adaptive texture description based neighbourhood segmentation method is studied in the paper, the major and minor relationship of the impacts the probabilistic adaptive texture model and the neighbourhood segmentation method has respectively on the segmentation is analyzed through the experiment. The experimental result shows that it is the neighbourhood based segmentation method which mainly contributes to the good performance （with less than 1.34% error rate） whereas the probabilistic model contributes less. This conclusion is helpful for method＇s improvement.