用于纹理特征提取的改进的成对旋转不变共生局部二值模式算法

发布时间：2018-05-07 02:43

本文选题：特征提取 + 局部二值模式　；参考：《计算机应用》2016年12期

【摘要】：针对用于纹理特征提取的成对旋转不变共生局部二值模式(PRICoLBP)算法计算特征维度大、旋转不变性较差、对光照变化敏感的问题,提出一种融合局部纹理信息的改进PRICoLBP算法。首先,分别最大化和最小化图像像素点的二值序列,得到两个邻域像素点的坐标,由中心像素点坐标和得到的邻域像素点坐标计算出共生点对的坐标;其次,利用完备二值模式(CLBP)算法提取图像的每个像素点的纹理信息。在相同分类器下,对Brodatz、Outex(TC10,TC12)、Outex(TC14)、CURe T和KTH_TIPS数据库的分类实验中,所提算法的识别率比PRICoLBP算法分别提高了0.17、0.24、2.65、2.39和2.04个百分点。实验结果表明,所提算法在处理纹理旋转变化、光照条件多样的图像时具有较好的识别效果。
[Abstract]:Aiming at the problem that the pairwise rotation invariant local binary pattern (PRICoLBP) algorithm for texture feature extraction is characterized by large feature dimension, poor rotation invariance and sensitivity to illumination changes, an improved PRICoLBP algorithm is proposed to fuse local texture information. Firstly, the binary sequence of pixel points in the image is maximized and minimized respectively, and the coordinates of the two neighboring pixels are obtained, and the coordinates of the symbiotic points are calculated from the coordinates of the central pixel points and the resulting neighborhood pixels. The texture information of each pixel of the image is extracted by the complete binary mode CLBP algorithm. Under the same classifier, the recognition rate of the proposed algorithm is 0.17% 0.242.65% and 2.04% higher than that of the PRICoLBP algorithm in the classification experiment of Brodatz's outlier TC10 / TC12 / TC14 / ure T and KTH_TIPS database, respectively. The experimental results show that the proposed algorithm has a good recognition effect when dealing with images with varying texture rotation and various illumination conditions.
【作者单位】：西南交通大学机械工程学院;
【基金】：国家自然科学基金资助项目(5127543) 四川省科技支撑计划项目(2015GZ0200)~~
【分类号】：TP391.41

【相似文献】