摘要
根据真实图像和计算机生成图像在采集方式上的不同所导致的关联性差异,建立了基于隐马尔科夫模型和图像能量谱特性的插值检测算法,选用支持向量机作为分类器进行训练和测试,对真实图像和计算机生成图像进行了分类。在哥伦比亚大学真实图像和计算机生成图像数据库上的实验结果表明,该方法可以达到98.8%的准确率,具有广阔的应用前景。
A new computer graphics detection scheme based on interpolation detection was proposed. Hidden Markov model (HMM) and image energy spectrum were used in extracted features. Kernel-based support vector machine (SVM) was chosen as a classifier to train and test the given image. Experimental results on Columbia Photographic Images and Photorealistic Computer Graphics Dataset demonstrated that the detection rate reached 98.8%, and the proposed approach possessed promising capability in identifying computer graphics.
出处
《测绘科学技术学报》
北大核心
2009年第4期288-291,共4页
Journal of Geomatics Science and Technology
基金
国家自然科学基金项目资助(60473022)
关键词
真实图像
计算机生成图像
插值检测
隐马尔科夫模型
图像能量谱
支持向量机
natural image
computer graphics (CG)
interpolation detection
hidden Markov model (HMM)
image energy spectrum
support vector machine (SVM)