LBP纹理特征提取算法。首先将检测窗口划分为16×16的小区域（cell），对于每个cell中的一个像素，将其环形邻域内的8个点（也可以是环形邻域多个点，如图 3‑ 4. 应用LBP算法的三个邻域示例所示）进行顺时针或逆时针的比较，如果中心像素值比该邻点大，则将邻点赋值为1，否则赋值为0，这样每个点都会获得一个8位二进制数（通常转换为十进制数）。然后计算每个cell的直方图，即每个数字（假定是十进制数）出现的频率（也就是一个关于每一个像素点是否比邻域内点大的一个二进制序列进行统计），然后对该直方图进行归一化处理。最后将得到的每个cell的统计直方图进行连接，就得到了整幅图的LBP纹理特征-LBP texture feature extraction algorithm. First detection window is divided into 1616 small area (cell), one pixel in each cell, its circular neighborhood of 8 points (You can also ring neighborhood, as shown in Figure 3-4 example of application of the LBP algorithm, three neighborhood shown) clockwise or counterclockwise, the center pixel value than the adjacent points, then adjacent points assigned to 1, otherwise the assignment is 0, so that each point will an 8-bit binary number (usually converted to a decimal number). And then computing the histogram of each cell, each number (assuming that is a decimal number) the frequency (that is, one on each pixel next to the interior point for a binary sequence of statistics), then the histogram normalized. Finally, get the histogram of each cell to connect, to get the whole image of the LBP texture features