中国科学院机构知识库网格
Chinese Academy of Sciences Institutional Repositories Grid
CR image filter methods research based on wavelet-domain hidden markov models (EI CONFERENCE)

文献类型:会议论文

作者Wang J.-L.; Wang J.-L.; Li D.-Y.; Wang Y.-P.
出版日期2006
会议名称ICO20: Optical Information Processing, August 21, 2005 - August 26, 2005
会议地点Changchun, China
关键词In the procedure of computed radiography imaging we should firstly get across the characters of kinds of noises and the relationship between the image signals and noises. Based on the specialties of computed radiography (CR) images and medical image processing we have study the filtering methods for computed radiography images noises. On the base of analyzing computed radiography imaging system in detail the author think that the major two noises are Gaussian white noise and Poisson noise. Then the different relationship of between two kinds of noises and signal were studied completely. By considering both the characteristics of computed radiography images and the statistical features of wavelet transformed images a multiscale image filtering algorithm which based on two-state hidden markov model (HMM) and mixture Gaussian statistical model has been used to decrease the Gaussian white noise in computed images. By using EM (Expectation Maximization) algorithm to estimate noise coefficients in each scale and obtain power spectrum matrix then this carried through the syncretized two Filter that are IIR(infinite impulse response) Wiener Filter and HMM according to scale size and achieve the experiments as well as the comparison with other denoising methods were presented at last.
收录类别EI
源URL[http://ir.ciomp.ac.cn/handle/181722/33623]  
专题长春光学精密机械与物理研究所_中科院长春光机所知识产出_会议论文
推荐引用方式
GB/T 7714
Wang J.-L.,Wang J.-L.,Li D.-Y.,et al. CR image filter methods research based on wavelet-domain hidden markov models (EI CONFERENCE)[C]. 见:ICO20: Optical Information Processing, August 21, 2005 - August 26, 2005. Changchun, China.

入库方式: OAI收割

来源:长春光学精密机械与物理研究所

浏览0
下载0
收藏0
其他版本

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。