| 摘要 |
A problem inherent to radiographic images, which may occur when an independent component analysis technique is applied to energy subtraction carried out on radiographic images, is solved to achieve separation of image components to be separated with higher accuracy. As preprocessing before the independent component analysis, a spatial frequency band which contains the components to be separated is extracted, pixels of the radiographic images are classified into more than one subsets for each radiographic image based on a value of a predetermined parameter, and/or nonlinear pixel value conversion is applied to the radiographic images based on a value of the predetermined parameter. Alternatively, nonlinear independent component analysis is carried out according to a model using the predetermined parameter.
|
