基于邻域非一致性图像过渡区提取与分割  

Image transition region extraction and segmentation based on neighborhood unhomogeneity

作　　者：王彦春[1] 梁德群[1] 王演[1] 王国良[1] 

机构地区：[1]大连海事大学信息工程学院,辽宁大连116026

出　　处：《光电子．激光》2008年第3期404-408,共5页Journal of Optoelectronics·Laser

基　　金：交通部科学技术资金资助项目(200136422504)

摘　　要：通过对图像过渡区像素属性的深入分析,提出基于邻域非一致性的过渡区直接提取算法。邻域非一致性的滤波作用提高了算法的抗噪性。摆脱了对受剪切值Llow与Lhigh的依赖。实验结果表明,本文算法优于传统的基于梯度算子的过渡区间接提取方法和基于局部复杂度的直接提取方法。The conventional transition region extraction methods are based on gradient operator,which are sensitive to noise and restricted by Llow and Lhigh. By analyzing pixel properties of transition regions,a neighborhood unhomogeneity-based transition region extraction method （NU-TREM）, a direct method to extract transition regions,was presented,The filtering ability of neighborhood unhomogeneity improves the ability of U-TREM to deal with noise,U-TREM is depending no more on Llow and Lhigh. Experimental results demonstrate that U-TREM significantly outperforms the conventional gradient-based transition region extraction methods （G-TREM） and the local complexity-based transitional region extraction method （C-TREM）.

关 键 词：图像提取与分割 邻域非一致性 过渡区 阈值 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]