参数自适应的耦合卡通-纹理分解及边缘检测方法  

Adaptive parameters approach for combining image decomposition with contour detection

作　　者：张力娜[1,2] 李小林[2] 

机构地区：[1]咸阳师范学院数学与信息科学学院,陕西咸阳712000 [2]咸阳师范学院图形图像处理研究所,陕西咸阳712000

出　　处：《激光与红外》2012年第8期940-943,共4页Laser & Infrared

基　　金：陕西省自然科学基础研究计划项目(No.2011JE011);陕西省教育厅专项基金项目(No.11JK1050);咸阳师范学院专项科研基金项目(No.11XSYK303;09XSYK303)资助

摘　　要：在耦合卡通-纹理分解及边缘检测模型的基础上,提出了根据原图像在不同位置的图像特征自适应选取参数的算法,即根据对原图像预分析的结果,分别在边缘、纹理、分片光滑区域选取不同的参数值,从而达到充分提取纹理、保证卡通图像光滑性同时保护边缘的目的。数值实验结果表明,本文方法提高了图像分解的质量及边缘检测的准确性,减少了边缘提取的奇异点。Based on the model that combines image decomposition with contour detection, it is put forward an adaptive method that select the most appropriate parameters adaptively according to the image feature in different positions of the original image. Based on the results of the pre-analysis of original image, different parameters are selected respec- tively in the edge, texture, piecewise smooth area. That achieves the full extraction of texture, guarantees cartoon image smoothness while protecting the edges of the objective. Numerical experiments show that the proposed method improves the contour detection accuracy, reduces the singularity and makes the cartoon-texture decomposition more thoroughly.

关 键 词：图像分解 边缘检测 自适应参数 卡通 纹理 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]