中文说明:纹理分析的图像分析非平凡的一部份。不同的纹理图像的分类必须旋转等各种参数的鲁棒性。在此代码中使用的解决方案是局部二进制模式 (LBP) 修改具有旋转不变性,因为它代表了复杂性和性能之间最好的折衷办法。纹理分析应该是理想的情况是不变的观点。一种好的纹理分类算法时,必须考虑到几个方面。他们当中包括照明、 缩放、 旋转和投影不变性、 与噪声的鲁棒性和其他参数,以及窗口大小要求和计算复杂度。考虑到的只有不变方面是旋转和缩放不变性和计算复杂度。该解决方案已经过测试的纹理数据库。13 纹理的是树皮砖泡沫草皮革猪皮椰砂稻草水编织木材和羊毛。有 7 版本的纹理在每个类中,对应于 7 不同旋转角度: 0 °、 30 °、 60 °、 90 °、 120 °、 150 °、 200 °。结果与最近邻居 calssifier 和 k-折交叉验证每个纹理被划分成 4 个区域,生成更多的积分 98.9%正确分类的原始数据集和 93%。Test1 文件夹包括用于旋转不变的 LBP 特征的代码。Test2 文件夹包含代码的原始 Brodatz databasetest 3 文件夹分类包括分类分区数据库的代码旋转的纹理图像的纹理数据库都可免费在这里: http://sipi.usc.edu/database/database.php?volume=rotate
English Description:
Texture analysis represents a non-trivial part of image analysis. The classification of different textures images have to be robust to various parameters such as rotation. The solution used in this code is Local Binary Pattern (LBP) modified for rotation invariance because it represents the best compromise between complexity and performance.Texture analysis should be ideally invariant to viewpoints. A good texture classification algorithm must take into account several aspects. Among them are illumination, scaling, rotation and projection invariance, robustness with relation to noise and other parameters as well as window size requirements and computational complexity. The only invariant aspects taken into account are rotation and scaling invariance and computational complexity.The solution has been tested on the Brodatz database.The 13 textures are Bark, Brick, Bubbles, Grass, Leather, Pigskin, Raffia, Sand, Straw, Water, Weave, Wood a