# Find Maxima vs. Adjust Threshold

The ImageJ user guide states:

"[Find Maxima] Determines the local maxima in an image and creates a binary (mask-like) image of the same size with the maxima, or one segmented particle per maximum, marked [29]. Analysis is performed on the existing rectangular selection or on the entire image if no selection is present. […]

Noise Tolerance: Maxima are ignored if they do not stand out from the surroundings by more than this value (calibrated units for calibrated images). In other words, a threshold is set at the maximum value minus noise tolerance and the contiguous area around the maximum above the threshold is analyzed. For accepting a maximum, this area must not contain any point with a value higher than the maximum. Only one maximum within this area is accepted."

I’m interested in the output “Maxima within Tolerance”, which is: “All points within the Noise Tolerance for each maximum.”

Can anyone help me understand how this is different from binarizing a photo using the Threshold function?

In the end I will use the Analyze Particles function to count the number of cells in a photograph. It seems like Find Maxima and Adjust/Threshold are two paths to the same place.

When would I use one, and when the other? I’m having trouble understanding which is best for me and it isn’t yet clear in experimenting what provides better results.

(Threshold: “Use this tool to automatically or interactively set lower and upper threshold values, segmenting grayscale images into features of interest and background.”)

Hermes

Hi @Hermes,

The short answer is that Find Maxima gives points, and Threshold gives you regions.

• Find Maxima gives you only a set of points. This is convenient for counting, especially if the cells tend to cluster. It works if you can tweak the parameters to reliably get exactly one local maximum point per cell, and none in the background. In my experience, some cell stains don’t work well with this method, possibly because they have brighter and dimmer areas within a cell. In that case the threshold method might be better.

• Threshold gives you the actual areas that are above the threshold. To count the underlying objects, you need to do Analyze > Analyze Particles to define connected regions. If you have clusters, though, you need to process further to separate the connected cells, so you can get an accurate count. The thresholded regions can be used to get more information about the cells, e.g. signal intensity, area, or shape.

If they both work for your images, then it’s all good!

Hope this helps.

1 Like