| 摘要 |
There is provided an image processing method using a novel filter bank based on a wavelet analysis for distinguishing with a high accuracy, circular independent abnormal shades and linear abnormal shades of different sizes contained in a medical image such as a mammogram from a normal organ such as a blood vessel and a rib. In order to detect different circular components and/or linear components contained in the input image signal, the filter bank includes: a division section having a high-pass filter given by a transfer function of HH(Z) = (-Z+Z<-1>)/2 and a low-pass filter given by a transfer function of HL(Z) = (Z+Z<-1>)/2; and a synthesis section having a high-pass filter given by a transfer function of FH(Z) = (Z-Z<-1>)/2 and a low-pass filter given by a transfer function of FL(Z) = (Z+Z<-1>)/2. The minimum unique value and/or the maximum unique value of the Hessian matrix having, as elements, an image signal divided into respective bands by the division section are used when generating the output image signal. |
