基于多层理想滤波器的纹理特征提取方法

Extraction Method of Texture Features Based on Multi-layer Ideal Filters

提出一种利用多层理想滤波器进行纹理特征提取的新方法,该方法克服了传统Gabor滤波器的带宽限制,避开了Gabor滤波器和Log Gabor滤波器构造过程中的复杂的参数计算,在保持同等性能的前提下,简化了问题。对多种纹理进行的特征提取和分割实验表明,该方法对纹理特征的捕获能力与Gabor、Log Gabor方法基本相当,提高了分割效率。

This paper proposes a new method for extracting texture features by using multi-layer ideal filters. The method overcomes the bandwidth limitation of traditional Gabor filters, and avoids the complexity of parameters computing in procedure of constructing Gabor filters or Log Gabor filters. On the premise of keeping same performance, the new method predigests the problem. Experiments to extract features and segment multi kinds of textures show that, the new method has almost the same capability in capturing texture features as Gabor and Log Gabor, and improves the efficiency of segmentation.