19th Dec 2023: Comparing different masking methods

Automatically finding a mask/ROI of the wells is key to subsequently computing parameters like Fv/Fm. In this page a couple of different masking approaches are compared.

1. Methods

• All methods are based on a thresholding approach
• We vary a parameter called num_std, where the threshold is set as the mean of a blank well + some multiple of the standard deviation of the intensities in the blank well: mu + num_std * sigma
• We experiment with morphological opening, to smooth the mask and remove noise
• In order to compare a well image (with many time points) to a single threshold value, we experiment with using the pixel-wise mean and minimum

2. Results

The following plots show metrics related to the masks, where the x-axis is the plate (arbitrary order).

Mean reduction, no opening:

Min reduction, no opening:

Min reduction, with opening:

3. Conclusions

• The default value (min reduction, sigma=3) found already looks fairly optimal to me
• Average Fv/Fm values are fairly insensitive to this, which is good