Phase Unwrapping Problems in Digital Holographic Microscopy of Biological Cells - Biomedical Optical Phase Microscopy and Nanoscopy - page 130

(A)

(B)

Figure 7.4
USAF resolution target (A) without curvature correction and (B) with curvature correction
applied. The images are 174 μ m 174 μ m (450 450 pixels).

3 3

shapes. In order to separate the background from the cell, an algorithm, based on modified
polynomial fitting [7] , is implemented. The method goes as follows:

1. Fit a polynomial surface to the unwrapped phase image.
2. Compare the polynomial surface with the original image pixel by pixel and retain the
lower of two values. Generate a new profile.
3. Substitute the unwrapped phase image with the new profile obtained in step 2.
4. Repeat steps 1, 2, and 3 several times, until the profile converges.

Finally, the processed background is subtracted from the original phase image, yielding a
phase profile with a near-null background. This method, however, while substantially
reducing the background, tends to “overcompensate” any curvature present in the original
unwrapped phase image. To illustrate this point, consider a curved profile ( Figure 7.6 ). The