基于改进LBP的纹理特征提取算法

Texture Feature Extraction Algorithm Based on Improved LBP

LBP纹理特征提取算法在提取纹理特征时,存在鲁棒性较差、对噪声较敏感等问题。针对前述问题,文章提出一种(ILNRBP)改进的局部抗噪鲁棒性二值模式。对于一个给定的中心像素,首先计算中心像素灰度值与邻域像素点周围的四个邻域像素点灰度均值之间的差值,然后将该差值与阈值T之间的差异二值量化得到ILNRBP二进制串,最后根据所有像素的ILNRBP直方图得到该幅图像的特征直方图。

LBP texture feature extraction algorithm has the problems of poor robustness and sensitiveness to noise when extracting texture features.Considering the above problems,an Improved Local Noise Robustness Binary Pattern (ILNRBP) is proposed.For a given center pixel,firstly,the difference value between the central pixel gray degree value and the pixel gray degree mean of the four neighborhood pixels around the neighborhood pixels is calculated.Then the difference between the difference value and threshold T is quantized to obtain the ILNRBP binary string.Finally,the characteristic histogram of the image is obtained from ILNRBP histogram according to the all pixels.