Automatic Equal Areas Method

Original Image (binarized)

Original (input) Image

The circle overlaid on the Input Image shows where we are going to crop the image, since we want to avoid the reflection coming off the dish. That reflection will really throw off the sensitivity optimization.

Every pixel in the right (binarized) image is either pure white or pure black, with nothing in between.

Original Image (cropped and binarized)

Blank plot. The area of this circle is equal to the area of the Petri dish.

Generated Matlab Plot (filled white circle plotted everywhere a colony was detected and everything outside the circle was set to black)

Every time our function detected a new colony a white filled circle was plotted on top of the Blank Plot with the same position and radius as the real colony.

The light area in the Generated Matlab Plot is approximately equal to the light area in the cropped and binarized version of the Original Image.

The Equal Areas method determined the optimal Sensitivity to be 0.8852 and 103 colonies were detected.

No false positives were detected, but it failed to detect 5 intergrown colonies. Note that this is a slight underestimate since when we cropped the dish to avoid the reflections, we also cut off a few small colonies that had grown near the perimeter (though if we hadn’t done that we would have ended up with a gross overestimate).