Figure 1.
Overview of PombeX cell segmentation method.
Figure 2.
Analysis of pixel intensity variation of the same specimen imaged at different z-focus position.
Top plot: original image at one z-position with red contours indicating k-means cluster boundaries. Middle plot: the same image with color-coded cluster membership for pixels. Bottom plot: Mean intensity of each pixel cluster through z-position.
Figure 3.
Examples of images before (left) and after (right) adaptive focus gradient correction and contrast enhancement.
In the left panel the original images are overlaid with nucleus centroids (black and white spots) and the estimated z = 0 line (green line, where specimen plane intersects focal plane). For the nucleus centroids, black means its average intensity is lower than background, and white means it is brighter than background.
Figure 4.
Example results of distance transform-based procedure.
Original images (left), pixel classification before (middle) and after (right) distance transform-based procedure. Blue pixels indicate background, red is cell membrane and green is cell interior region.
Figure 5.
Representative segmentation results on various pombe mutants.
For each image pair, trans-illumination images with randomly colored cell contours are shown on the left, and the corresponding RPA-CFP images on the right.
Figure 6.
Segmentation results on well-known phenotypes.
Cell contours are colored randomly.
Figure 7.
Segmentation method comparison.
Contours of different cells are colored randomly.
Table 1.
Quantitative performance results of yeast segmentation algorithms.