摘要
纹理粗糙度是纹理最高层次的视觉感知特征,对底层图像特征向高层语义特征转化具有最要意义,为此,提出了大量的度量纹理粗糙度的算法,但是这些方法往往是基于不同的应用背景,没有总体比较和评价,多局限于灰度共生矩阵的选用和改进,算法普遍存在计算量过大、误差较大、应用能力模糊的缺点。基于常用的5种度量纹理粗糙度的算法,在不同图像源以及噪声图像上进行实验,测试不同度量方法所具有的纹理分辨能力、旋转不变性以及算法鲁棒性,进而给出选择度量纹理粗糙度算法的参考模型。
Texture coarseness is the highest level of texture feature of visual perception.It is very important that lower features are transformed to higher semantic features.Many measure texture coarseness algorithms of image were proposed based on different application background.But they are lack of comparison and evaluation of the overall and are only limited to the selection of the GLCM method.This paper made a testing among these methods in different images and noise images.The experimental results show that five kinds of texture coarseness measurement algorithm have different textural performance,rotation invariance and algorithm robustness.A reference model was given to select diffe-rent algorithm of texture coarseness measurement algorithm based on different application.
出处
《计算机科学》
CSCD
北大核心
2011年第6期288-292,共5页
Computer Science
基金
国家自然科学基金项目(60863010)
973前期计划专项课题项目(2010CB334709)
新疆自然科学基金项目(2010211a19)资助
关键词
纹理粗糙度
纹理分辨能力
旋转不变性
鲁棒性
Textural coarseness
Textural performance
Rotation invariance
Robustness
