This step classifies the femur and patellar. This method has
the stream from the raw (Fig. 3(a)) to the binarization (Fig.
3(b)), the closing (Fig. 3(c)), and the extraction (Fig. 3(d)).
Since the intensity of border between the soft tissue and the
bone (cancellous and cortical bone) region in the raw MDCT
image has fuzzy distribution (Fig. 3(a)), to set the threshold of
the binarization by manual is difficult, has not quantitative,
and has take a lot of hard works. Therefore, it is important to
binarize the raw image using the accurate and automatic
threshold. If threshold is too small and large, it is impossible to
segment the femur and patellar. This process secondary
differentiates the intensity-frequency graph, detects the
maximum value of the threshold to segment the femur and
patellar dynamically. Fig. 4(a) indicates the frequency
(number) of the intensity for the object pixel, and has the
intensity between the soft tissue and bone about 300. Fig. 4(b)
indicates the secondary differentiation for the intensity to
frequency, and determines the threshold of the intensity
between the soft tissue and bone about 300.