The Application of Cellular Automaton in Medical Semiautomatic Segmentation - Cellular Automata in Image Processing and Geometry - page 166

Fig. 9.3 Single and multiple objects segmentation. First column: original images. Second
column: initial segmentation and the input markers (the green and blue lines indicate the
object and background respectively). Last column: segmentation results by the proposed RCA
method.
Fig. 9.4 First column: original images. Second column: initial segmentation and the input
markers. Third column: segmentation results by the MSRM method. Last column: segmen-
tation results by the proposed RCA method.
image, which implies that more markers may be helpful to provide accuracy results.
To quantitatively evaluate RCA method, the desired objects are manually seg-
mented as ground-truth and three indicators: running time, true positive rate (TPR)
and false positive rate (FPR) [25] are computed on the test images. Table 9.1 lists
the comparison between two methods with same user markers and programming en-
vironment. The numeral results demonstrate that RCA achieves the better segmen-
tation performance. Experiments prove that the time consuming of RCA depends
on initial segmentation result, user markers and image content.