期刊名称：International Journal of Advanced Computer Science and Applications(IJACSA)
印刷版ISSN：2158-107X
电子版ISSN：2156-5570
出版年度：2018
卷号：9
期号：4
DOI：10.14569/IJACSA.2018.090459
出版社：Science and Information Society (SAI)
摘要：Iterative image reconstruction methods are considered better as compared to the analytical reconstruction methods in terms of their noise characteristics and quantification ability. Penalized-Likelihood Expectation Maximization (PLEM) image reconstruction methods are able to incorporate prior information about the object being imaged and have flexibility to include various prior functions which are based on different image descriptions. Median Root Priors intrinsically take into account the salient image features, such as edges, which becomes smooth owing to quadratic priors. Generally, a 3*3 pixels neighborhood support or root image size is used to evaluate the median. We evaluate different root image sizes to observe their effect on the final reconstructed image. Our results show that at higher parameter values, root image size has pronounced effect on different image quality parameters evaluated such as reconstructed image bias as compared to the phantom image, contrast and resolution in the reconstructed object. Our results show that for the small-sized objects, small root image is better whereas for objects of diameter more than two to three times of the resolution of the reconstructed object, larger root image size is preferable in terms of reconstruction speed and image quality.
关键词：Penalized-Likelihood expectation maximization; median root priors; maximum-likelihood expectation maximization; full-width-at-half-maximum
Loading...
