基于卡尔曼滤波SVM图像鉴别方法研究  

Simulation of Image Identification Based on Kalman and SVM

作　　者：周立波[1,2] 刘合安[2] 

机构地区：[1]湖南大学,湖南长沙410000 [2]湖南城市学院,湖南益阳413000

出　　处：《计算机仿真》2012年第3期316-318,322,共4页Computer Simulation

基　　金：国家环保公益基金(201009022)

摘　　要：研究如何区分自然图像和计算机图形判别图像问题。由于人眼难以判别真伪图像,针对传统的图像鉴别方法由于参数的设置而造成识别准确率低,从而无法准确的区分自然图像和计算机图形。提出了一种卡尔曼滤波和支持向量机分类算法相结合的高性能图像真伪鉴别方法。算法首先采用卡尔曼滤波器提取图像的各个级别的分量,求解出各级分量和判别的误差值并作为特征数据,然后采用支持向量机分类算法对特征数据进行训练判别。仿真结果表明,提出的算法可以有效的判别图像的真伪,可以有效的识别出原始自然图像和合成图像,识别率高达99%以上,同时算法的稳定性较高,具有一定的实际应用价值。The research deals with the distinguish of natural images and synthetic images and the discriminant of image authenticity.The traditional image identification method can not accurately distinguish natural images and computer graphs.In order to solve the above problems,we put forward a high-performance image identification algorithm method based on kalman filter and support vector machine.The algorithm firstly extracted every level of the components of the image by using kalman filter,worked out the various components and discriminant error as the characteristic data.Then,support vector machine was used to train and discriminate feature data classification algorithm.Simulation results show that the proposed algorithm can effectively discriminate the image authenticity,effectively identify original natural images and synthetic images,and the recognition rate is 95% or higher.

关 键 词：自然图像 卡尔曼滤波 特征提取 支持向量机 人造图像 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]