基于形状索引的加权局部凹微结构模式及其纹理分类  

Texture Classification Based on Shape-Indexed Weighted Local Concave Microstructure Patterns

作　　者：吴玉欢 陈熙 WU Yu-huan;CHEN Xi(Guizhou Normal University College of Big Data and Computer Science,Guiyang 550025,China)

机构地区：[1]贵州师范大学大数据与计算机科学学院,贵州贵阳550025

出　　处：《电脑与电信》2024年第11期17-23,共7页Computer & Telecommunication

基　　金：国家自然科学基金资助(61762022);湖南省教育科学规划课题“基于VR技术的高中物理沉浸式教学的研究与实践”(XJK23BJC011)。

摘　　要：大多数基于局部二值模式的方法受不同条件下由于局部区域的像素变化导致的特征改变,难以有效区分结构相似但对比度不同的局部区域以及对噪声敏感的缺陷。为了解决这些问题,提出了一种新的局部凹微结构模式,即基于形状索引的加权局部凹微结构模式(Shape-Indexed Weighted Local Concave Microstructure Pattern,SIWLCvMSP)。该算法首先提取图像的形状指数,然后对形状指数计算局部加权凹微结构模式作为图像的特征。形状是图像一种稳定特征,而形状指数是一种描述形状的有效方法,基于形状指数的算法能够在噪声环境下更有效地提取图像的判别特征,所提取的特征更具有稳定性和判别性。该方法分别在五个广泛使用的图像数据库上进行大量实验,结果表明,SI-WLCvMSP描述符可以提高分类精度以及对噪声具有抗敏感性。Most methods based on Local Binary Patterns are affected by feature changes due to pixel variations in local regions under different conditions,making it difficult to effectively distinguish between locally similar structures with different contrasts and to address the sensitivity to noise.To tackle these issues,a new Local Concave Microstructure Pattern is proposed,namely the Shape-Indexed Weighted Local Concave Microstructure Pattern(SI-WLCvMSP).This algorithm first extracts the shape index of the image,and then computes the local weighted concave microstructure pattern based on the shape index as the image feature.The shape serves as a stable feature of the image,while the shape index is an effective method for describing shape.Algorithms based on the shape index can more effectively extract discriminative features of the image in noisy environments,resulting in features that are more stable and discriminative.Extensive experiments conducted on five widely used image databases demonstrate that the SI-WLCvMSP descriptor can improve classification accuracy and exhibits robustness against noise sensitivity.

关 键 词：形状指数 加权局部凹微结构模式 纹理分类 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]