基于Contourlet特征修正的纹理图像识别算法

Texture Image Recognition Algorithm Based on Contourlet Characteristic Modification

提出一种基于Contourlet特征修正的纹理识别算法,不同分辨率下采取不同的方法提取特征,根据统计规律对每个方向上的纹理特征进行修正增强,利用支持向量机进行识别,以提高纹理图像识别的准确性。实验证明,在受噪声干扰严重的情况下,该方法的识别正确率优于小波、小波包、Ridgelet等算法。

This paper proposes an algorithm based on Contourlet characteristic modification, which extracts the characteristics in different resolution with different methods, adjusts them by statistical analysis, and deals with the feature vectors by Support Vector Machines（SVM）. Compared with the existing transforms including Ridgelet, Wavelet and Wavelet package, the method consistently yields the best overall performance in recognition veracity rate, especially when the images are interfered by intensive noise.