From CT images having captured an abdomen, a two-dimensional characteristic space is assumed, in which X and Y axes indicate CT values of image data taken in first and second time phases different from each other. In this space, a two-dimensional histogram concerning respective pixels located at the same position in the two time phases is taken as a sample distribution of pixels. A two-variable distribution function is applied to this sample distribution, so as to estimate a matrix distribution of pixels corresponding to the whole liver region. According to this matrix distribution, a range of CT value of pixels corresponding to the liver region is estimated.