医学CT图像预处理上下文算法研究  

An Preprocessing Algorithm of Medical CT Image With The Context Model

作　　者：史延新[1] SHI Yanxin(School of Science,Xi an Technology University,Xi an 710032,China)

机构地区：[1]西安工业大学理学院,陕西省西安市710032

出　　处：《电子科技》2018年第9期72-76,共5页Electronic Science and Technology

基　　金：陕西省教育厅专项科研计划项目(14JK1345)

摘　　要：医学CT图像的纹理特征差异包含着重要的医疗信息,在预处理时通常会和噪声一起被处理。为了保护纹理特征,通过对大量医学CT图像的研究,选择灰度均值、熵以及分形维数和分形截距共同构成特征向量作为纹理特征,并引入上下文模型获取上下文信息,使用LBG算法进行上下文量化,进行医学CT图像预处理。实验结果表明本文算法能够在提高图像质量的同时较好地保持纹理等细节信息,取得良好的视觉效果,峰值信噪比和加权的结构相似性指数也优于现有算法。The difference in texture features is very important in medical CT images. It usually be smoothed as the noise while preprocessing. In order to preserve the texture features, a large number of medical CT images has been studied. The feature vector is formed as texture features composed with the mean of the gray scale, entropy, fractal dimension and fractal intercept. The context model is introduced to obtain the context for preprocessing medical CT images also. LBG algorithm is used to quantify context to preprocess medical CT images. The experimental result shows that the proposed algorithm can preserve the texture details better to improve the quality of the medical CT images. The subjective evaluation shows good visual effect. The objective evaluation with the peak signal to noise ratio and the weighted structure similarity index is also outperformed conventional algorithms.

关 键 词：医学CT图像 纹理 熵 分形维 上下文 预处理 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]