摘要
为提高不同光照、不同角度条件下的纹理识别精度,提出了一种利用多级小波分解和多尺度旋转不变LBP融合的纹理提取算法。算法在传统的LBP特征提取基础上,采用多尺度的旋转不变LBP算子分别对多级小波逼近图像提取直方图序列特征向量,与各级小波能量进行加权融合,获取更多的纹理信息,对光照和角度的变化有更高的鲁棒性。仿真结果表明,相对传统的LBP特征提取算法,改进的算法具有更高的纹理识别率。
To improve the texture recognition accuracy under the condition of different illumination and angles, a texture extraction method by fusion of multi-level wavelet decomposition and LBP operator is proposed based on the traditional LBP feature extraction. The rotation invariant LBP algorithm which is multi-resolution was used to extract the Histogram sequence feature vector from the multi-level wavelet approximation image and each levels of wavelet energy weighted fusion for more information on the texture. Under the different illumination and angles it is higher robustness. The emulation show that improved texture algorithm have higher recognition rate compared with traditional algorithm.
出处
《微计算机信息》
2012年第3期135-136,170,共3页
Control & Automation
关键词
纹理识别
LBP算子
直方图序列特征向量
小波能量
texture recognition
LBP operator
histogram sequence feature vector
wavelet energy