Web图像清洗技术的研究与实现

RESEARCH AND IMPLEMENTATION OF WEB IMAGE CLEANING

目前 Web文档中充斥着各种图像 ,因此 ,对 Web上的图像进行清洗就显得非常必要 .提出了 Web图像清洗的概念 ,给出了一种基于机器学习的 Web图像清洗方案 ,并实现了基于此方案的系统 .将图像特征的提取建立在 Web文档的 DOM结构之上 ,并在决策树算法的基础上对多种特征表示及组合进行了实验和评估 .实验结果表明 ,提出的 Web图像清洗方案切实可行 ,清洗方法具有较快的速度和准确性 .

Now all kinds of images can be found in web documents, so it is necessary to clean web images. In this paper, the concept of web image cleaning is introduced, the framework of web image cleaning is designed, and a system of web image cleaning is implemented based on this framework. Image feature extraction is based on DOM tree, and experiment and evaluation of different feature combination are carried out based on the decision tree algorithm. According to the experiment result, the framework is feasible and the method works with fast speed and high accuracy.