Imaging and Confocal Tips

Imaging and Confocal Tips

Image Segmentation Thresholding

Fluorescent imaging starts with taking grayscale images of the different fluorphores that we have tagged to the proteins of interest in our samples. The pretty DAPI-FITC-TRITC images that we see in the journals all start out with 3 grayscale images that have been pseudo-colored blue, green or red. These images will have been captured using a grayscale range of 8-bit (0-255 gray levels separating black from white), 12-bit (0-4095 gray levels) or 16-bit (0-65,535 gray levels). If we want to measure something of interest in the image, we first have to segment the image into the things that we are interested in and the things that we aren't interested in.

Thresholding is the simplest method image segmentation in a grayscale image. It allows us to create a binary image showing what we are interested in and what we aren't based upon a level of brightness that we choose.


Here is a 16-bit image where the actin has been stained with Alexa Fluor 488.

We are going to apply a threshold of 750 to 65535 to the image.

In the graph above we can see the image histogram which graphs all of the pixel values from 0 to 65535 and we can see where we placed our limits at 750 and 65535. Displayed on our image in light blue, we can see where those pixels are.



  If we were interested in just the brighter objects we could raise the lower limit and look at the effect.

Here we have raised the lower limit to 1575.

In the image above we can see that fewer pixels are in the group with the floor at 1575.


If the objects that we are interested in aren't the dimmest or the brightest, we can set the top and bottom to reflect this:

Here we have selected the pixels that range from 358 to 962.

You can see in the image that we have selected not the dimmest pixels or the brightest pixels.



Finally, we may be interested in just the dimmest pixels in the image.


Here you can see that the threshold has been set to highlight the pixels that range from 0 or black to a gray level of 358.

Here you can see that the threshold has chosen all of the dark pixels.


Thresholding is usually the first process performed on a grayscale image to define the objects that we are interested in. From here, we can start to look at things like size or shape to further segment our image.