Документ взят из кэша поисковой машины. Адрес
оригинального документа
: http://imaging.cs.msu.ru/en/publication?id=264
Дата изменения: Sun Apr 10 03:41:18 2016
Дата индексирования: Sun Apr 10 03:41:19 2016
Кодировка: Windows-1251
Methods of noise filtering quality assessment for CT images
Publication type
Conference paper (abstract)
Author(s)
M.V. Storozhilova, A.S. Lukin, D.V. Yurin
Publication date
2013
Conference
Proceedings of 15-th International Conference "Digital Signal Processing and its Applications" (DSPA'2013)
Volume
2
Page(s)
88
Address
Moscow, Russia
Abstract
Noise filtering in computer tomography (CT) images plays an important role in reduction of radiation dose received by a patient. Building upon our previous work in noise reduction, this paper introduces two approaches for automatic evaluation of noise reduction quality. Both approaches are based on the analysis of residual images в?' differences between the original noisy image and the processed image. One method is computing the coefficient of local data correlation, while another one is using a concept of entropy to quantify randomness of the residual image. Both methods are evaluated on real-world CT scans and presented results suggest that they are effective for automatic selection of the noise reduction filtering strength parameter.
Comment
Noise filtering in computer tomography (CT) images plays an important role in reduction of radiation dose received by a patient. Building upon our previous work in noise reduction, this paper introduces two approaches for automatic evaluation of noise reduction quality. Both approaches are based on the analysis of residual images в?' differences between the original noisy image and the processed image. One method is computing the coefficient of local data correlation, while another one is using a concept of entropy to quantify randomness of the residual image. Both methods are evaluated on real-world CT scans and presented results suggest that they are effective for automatic selection of the noise reduction filtering strength parameter for both Rank-2.5D Ре NLM-2.5D filters. Comparison of the approaches based on autocorrelation and entropy show that for entropy measure graphs shapes are similar for both filtering algorithms differ from autocorrelation measure.
Appreciable effect appears even without block-wise algorithm application. The drawback of entropy approach is small relative maximum value (about 5%) differ from autocorrelation approach with relative maximum value about few times.
