说明：  细化参数的对数极坐标变换图像纹理特征提取算法提出一种细化参数的对数极坐标变换图像纹理特征提取算法,可以有效地消除旋转、缩放和平移等几何形 变的影响.在特征提取过程中,通过自相关图像消除平移的影响,引入细化参数的对数极坐标变换消除旋转和伸缩 的影响.在图像检索过程中,将不完全树型小波变换所得的特征矢量,通过欧式距离来度量图像之间的相似度.实 验表明,本算法对发生几何形变的纹理图像检索平均正确率达81.05 左右,较之传统算法能取得更好的检索 效果.
(This paper p ropo sed a new type of feat ure ext raction arit hmetic for text ure image,based on paramet rical log2polar t ranslation.This arit hmetic is robust to geo met ric distortions.During t he feat ure ext raction,t he autocorrelatio n image was int roduced to eliminate t he effect of t ranslatio n,while t he imp roved log2polar was used to eliminate t he effect s of rotation and scaling.In t he ret rieval p rocess,a set of inco mplete t ree st ruct ured wavelet t ransform of t he log2polar t ransformed autocorrelation image was used to measure similarity of text ure images.The experiment result s showed t hat t he average ret rieval rates of t his arit hmetic is abo ut 81.05 ,which is robust to t he geomet ric disto rtio ns as compared wit h t he typical algorit hms. )