Many thanks for your great comments.
I think I must write this subsection like the other subsections.
I am very eager to use figures and have more descriptions.
I apply all of your comments and inform you.

In machine learning, pooling is a technique to decrease the complexity of the images (feature maps).
We consider a set of pixels inside a squared window. The width of this window is determined by the poolsize.
We apply an operator like maximum (pool-max) on the selected pixels (the window).
In Gnuastro, there was no function to do this. So, I had to write and add this program to use in my own job.
I'm delighted to announce that the pooling program is now complete!
You can use any image as an input of this program. It means your input can be a rectangle or square image.
Also, it can contain blank values.
You must determine the poolsize (that can be any positive and non-zero value greater than the width of your
image) and the type of the desired operator(poolmax, pool-min, pool-mean, pool-median).
So if you test this program with the "during-dev.sh" script, you use the command below:

utilname=arithmetic
arguments="image.fits 2 pool-max"

That image is your desired input, 2 is the size of pooling, and "pool-max" is the selected operator.
I tested the program with different images and according to my understanding, it seems to be working correctly!
If you have any suggestions or comments, or if you encounter any issues while testing it with other images, please
let me know.