摘要： 分析了景物图像修补的特点，在此基础上构建了一类景物图像的样本数据集。尝试将支持向量方法引入该领域，采用支持向量机对待修补图像进行分类，利用匹配的模板数据对空洞进行边缘修补和色彩填充，根据空洞外部图像的细微特征对纹理和色彩进行二次修补。实验结果表明该方法的精度优于传统方法，体现了该方法的实用性和有效性。

关键词: 图像修补, 支持向量机, 模式识别, 边缘修补, 纹理修补

Abstract: This paper analyzes features of scenery image hole-filling manipulation. It builds a special image sample sets, classifies incomplete scenery image by support vector machie (SVM), filles color and texture in the blank hole, and modifies imperceptible color and texture. Theory analysis and experimental results show it has better applied value than several classical methods, such as structure inpainting, texture inpainting and image decomposition.

Key words: image hole-filling, support vector machine (SVM), pattern recognition, edge repair, texture repair

中图分类号: 

• TP391