数字式多媒体场景图像准确分类方法仿真  

Simulation of Accurate Classification Method for Digital Multimedia Scene Images

作　　者：许志强 张婷[2] XU Zhi-qiang;ZHANG Ting(School of Art & Design,Hubei University of Technology,Wuhan Hubei 430068,China;Engineering and Technology College,Hubei University of Technology,Wuhan Hubei 430068 ,China)

机构地区：[1]湖北工业大学艺术设计学院,湖北武汉430068 [2]湖北工业大学工程技术学院,湖北武汉430068

出　　处：《计算机仿真》2019年第7期385-388,共4页Computer Simulation

摘　　要：对数字式多媒体场景图像进行准确分类,能够有效管理海量的图像资源。对多媒体场景图像进行分类,需要估计多媒体场景图像分量,利用图像平均像素计算支持向量描述参数。传统方法对图像的语义信息进行融合,计算待分类图像至各聚类中心之间平均距离,但忽略了对多媒体场景图像分量的估计,导致分类精度低。提出基于SVDD的数字式多媒体场景图像准确分类方法。构建Gaussian-MRF模型,对图像分量进行估计,得到图像自身特性反射分量,同时利用增益补偿法恢复和校正图像亮度。利用异常目标集合尺寸与多媒体场景图像的空间分辨率,计算图像中异常目标的像元,利用像元和SVDD训练样本计算支持向量描述参数,根据所得描述参数完成多媒体场景图像的准确分类。实验表明,所提方法方法查全率约为98.5%,具有较高的分类精度,相较当前方法整体优越性较强,具备可实践性。To accurate classify digital multimedia scene images can effectively manage massive image resource. In this article, we focus on a method to accurately classify the digital multimedia scene images based on SVDD. At first, we built Gaussian-MRF model and estimated the image component to get the reflection component of image characteristic. Meanwhile, we used gain compensation method to recover and correct the image brightness, and then used the size of abnormal target set and the spatial resolution of multimedia scene image to calculate the image element of abnormal target in image. Moreover, we used the image element and SVDD training sample to calculate the support vector description parameter. Finally, we completed the accurate classification of multimedia scene images based on description parameters. Simulations show that the recall rate of proposed method is about 98. 5%, which has higher classification accuracy. Compared with the current method, the proposed method has strong overall superiority and practicality.

关 键 词：数字式 多媒体 场景图像 分类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]