ALGORITHM OF BINARY IMAGE PROCESSING - Fundamentals of Three-Dimensional Digital Image Processing - page 171

g ijk ← f ijk ,

u ijk ,
go to the next i .
(b-2) Otherwise go to the next i .
Perform the following procedures (2)

v ijk ←

(7), which are the same as the above
mentioned one, changing the directions of the coordinate axes and scanning
“(1) in the direction of the row (forward scan)” as is indicated below.
(2) Row direction (direction of the sux i ) backward scan
Input image:

∼

G , V

V
(3) Column direction (direction of the sux j ) forward scan
Input image:

G , Output image:

G , V

V
(4) Column direction (direction of the su x j ) backward scan
Input image:

G ,

Output image:

G , V

V
(5) Plane direction (direction of the sux k ) forward scan
Input image:

G ,

Output image:

G , V

V
(6) Plane direction (direction of the sux k ) backward scan
Input image:

Output image: G ,

G , V

G , V Output image:

[STEP 3] (Skeleton extraction)
Read a label existing in a label image V .Giveamarktoavoxelofalabel

image corresponding to the voxel readout here. Voxels marked here repre-

sent skeleton. A working table

U

T is used for search and retrieval of labels in

V .
(1) Generation of a work table for label retrieval.
Input image:

images U and

V = v ijk } { (label image)

Output: T = { t i } (work table) (1D array)

for all ( i, j, k ) do
t ( v ijk ) ←

1

enddo
(2) Generation of a skeleton image.
Input image: U = { u ijk } (label image), T (work table)

H { h ijk }

Output image: = (skeleton image)

for all ( i, j, k ) do
if t ( u ijk ) = 1
then h ijk ← 1

else h ijk ← 0

endif
enddo
When this algorithm is executed, three 3D arrays

can be
assigned to physically the same memory space unless an input image needs to

F , G ,and H