超像素词包模型与SVM分类的图像标注

Image annotation based on super-pixel bag of words model and SVM classification

为了改善基于词包模型与支持向量机(SVM)分类一幅图对应一个标签的单标签分类问题,提出了一种基于超像素词包模型与SVM分类的图像标注算法。将超像素分割结果作为词包模型的基本单元,用词包模型生成的视觉词汇表示超像素区域特征,保留了图像中的同质区域,很好地利用了图像的区域特征。仿真结果表明,该方法能有效改善基于词包模型与SVM分类的单标签分类问题,且分类的准确性有所提高。

In order to improve single label classification problem based on the bag of words model and SVM classification of a graph corresponding to a label,put forward an image annotation algorithm,which bases on superpixel bag of words model and SVM classification. The algorithm uses super-pixel segmentation results as basic unit of bag of words model,super-pixel regional characteristics expressed by visual words generated by bag of words model,keeps homogeneous area in image,take well use of regional characteristics of image. Simulation results show that the algorithm can effectively improve the single label classification problem based on bag of words model and SVM classification,and classification accuracy improved.