摘要
针对目前计算机生成图像鉴别算法在对图像纹理特征进行鉴别时精度较差的问题,提出一种基于长期控制计划(LTCP)特征的计算机生成图像鉴别算法.首先将彩色图像变换到颜色模型中,对图像进行下采样,获得较高尺度的纹理信息;然后采用基于LTCP特征和共生矩阵的计算机生成图像盲鉴别算法,对不同尺度的纹理图像LTCP特征及相邻像素一致性共生矩阵特征进行采集;最后通过判别分类器对LTCP特征及相邻像素一致性共生矩阵特征实施分类预测,根据分类预测结果实现计算机生成图像的鉴别.实验结果证明,该算法在计算机上生成的图像特征维度较低,鉴别率和精度较高,能实现计算机生成图像的准确鉴别.
Aiming at the problem that the current computer generated image identification algorithm had poor precision in image texture feature identification,the author proposed a computer generated image identification algorithm based on long-term control plan(LTCP)features.Firstly,the color image was transformed into the color model,and the image was sampled to obtain higher-scale texture information.Secondly,using the computer generated image blind identification algorithm based on LTCP features and co-occurrence matrix,LTCP features of texture images with different scales and adjacent pixel consistency co-occurrence matrix features were collected.Finally,the LTCP features and the adjacent pixel consistency co-occurrence matrix features were classified and predicted by discriminant classifier,and the computer generated image was identified according to the classification prediction results.The experimental results show that the proposed algorithm generates images on the computer with low feature dimension,high resolution and accuracy,which can accurately identify computer generated images.
作者
刘淑琴
LIU Shuqin(School of Computer Science & Technology,Xi’an University of Posts & Telecommunications,Xi’an 710121,China)
出处
《吉林大学学报(理学版)》
CAS
北大核心
2019年第2期393-398,共6页
Journal of Jilin University:Science Edition
基金
国家自然科学基金(批准号:61075014)
关键词
LTCP特征
计算机生成图像
鉴别算法
颜色空间
下采样
SVM分类器
long-term control plan(LTCP)feature
computer generated image
identification algorithm
color space
down sampling
SVM classifier