A virtual contour model of surface inhomogeneity is proposed in the paper aiming at measuring surface inhomogeneity of skin tumors. In addition, a novel feature cluster, including irregular and asymmetric descriptors of virtual contour, is constructed for measuring surface unevenness. The 2D surface unevenness of intensity distributions is converted to the irregularity of the virtual contour. Then features with irregularity and asymmetry are extracted. Furthermore, the features are selected and fused on Hausdorff distance metrics to implement the classification of skin tumors for the surface inhomogeneity using the irregularities and asymmetry of the virtual contours. Experimental results show that the model not only can achieve the dimension reduction, but also can improve the classifier performance.%为解决皮肤肿瘤表面不均匀性的度量问题,构建一种肿瘤表面灰度不均匀性的虚拟轮廓描述模型.提出虚拟轮廓不规则性和不对称性描述子,将肿瘤表面2D灰度分布的不均匀性转换为虚拟轮廓的不规则性,提取虚拟轮廓的不规则性和不对称性特征,依据Hausdorff距离进行特征选择和融合,实现基于虚拟轮廓不规则性和不对称性的肿瘤表面不均匀性分类.实验结果表明,该模型能实现数据降维,提高分类准确率.
展开▼
