基于轮廓波特征的纹理图像识别方法

通过利用方向可控滤波器(steerable filter)和轮廓波(contourlet)分解的方向性及能量变化特性,提出了一种适用于纹理图像的特征提取方法.利用基于轮廓波方向性与分解系数能量的性质可以很好地表示一个旋转纹理图像的特征,而且轮廓波的方向性和分解层次可以根据需要灵活地调节.使用这样的特征并运用简单的分类方法可以对旋转纹理图像进行较为准确的分类.最后通过对标准纹理库(Brodatz)图像的测试,验证了算法的有效性.

Contourlet-based feature extraction on texture classification

A new texture feature extraction method,which makes full use of the directional information and energy variation in steerable decomposition and contourlet decomposition,is presented.The feature vectors based on directional property and decomposition coefficients in contourlet decomposition can represent rotated texture's feature very well.According to different application purposes, the number of directional features and decomposition levels in contourlet decomposition can also be adapted easily.With the extracted feature vectors and a simple classification method,the high accurate classification rate can be reached with the samples from standard Brodatz texture database.The effectiveness of the algorithm is validated at last.