摘要
提出基于局部二值化模式和像素相关算子的半色调图像纹理特征提取方法,以实现误差分散类半色调图像的分类。该方法是将误差分散类图像先进行局部二值化模式变换,再以任一像素点为中心,取适当的距离提取八个方向的像素相关值作为图像的特征向量,最后将提取的特征通过BP神经网络进行分类。实验结果表明,提出的算法适用于二值图像的特征提取,能够降低局部二值模式的特征维数,提高时间效率和空间利用率;相对灰度共生矩阵算法提出的算法在计算复杂度、识别精度等性能方面都有所改善。
Feature extraction method was presented based on local binary pattern and pixel correlation operator for clas-sifying halftone images produced by various error diffusion methods.The error diffusion images transforms by using Ratio-LBP theory and takes eight directions pixel correlation value as a image feature vector from the appropriate distance around a pixel in the picture,and then it classifies the characteristics extracted by the BP neural network.The experimental results show the proposed methed can reduce the feature dimension of local binary pattern and improve time efficiency and space uti-lization when it is used to extract the feature parameters for a binary image.Compared with grey level co-cocurrence ma-trix,the proposed methed can improve the performance of image processing in computational complexity and recognition ac-curacy.
出处
《计算技术与自动化》
2015年第3期105-110,共6页
Computing Technology and Automation
基金
国家自然科学基金项目(61170102)
湖南省自然科学基金项目(14JJ2115
15JJ2046)
湖南省研究生科研创新项目(CX2015B566)
关键词
半调图像
局部二值模式
特征提取
像素相关算子
haftone image
local binary pattern
feature extraction
pixel correlation operator