Parallel Processing Strategies for Cell Motility and Shape Analysis - High-Throughput Image Reconstruction and Analysis - page 65

each pixel, its eight neighbors are assigned a value with the upper left as 0 and the
lower right as 128. A bit vector is formed with values from 0 to 255 by setting a
bit if the corresponding pixel is filled and clearing it if empty. Figure 3.13 shows 4
out of the 256 possible pixel arrangements. The value of the vector is used to index
a lookup table of all possible neighborhoods with the corresponding action. A 0
indicates the pixel should be cleared while a 1 leaves the pixel filled. The image
after thinning is applied still contains isolated loops which could be eliminated by
examining a larger neighborhood.

3.2.4 Results and Discussion
Figure 3.14 shows sample 4-level sets cell segmentation and GVD results. These
sample result corresponds to a wound healing image sequence, obtained from
phase-contrast microscopy, consisting of 136 frames of dimensions 300 × 300

Figure 3.14 Segmentation and tracking results for the wound sequence: (a) original image (frame
#135); (b) 4-level set mask + GVD; (c) neighborhood graph; and (d) cell trajectories.