If you have a Z-stack then there are some plugins implementing algorithms to find the most focused slice. But taking an unfocused image and de-blurring is not really available. The closest solutions would be deconvolution or machine-learning based methods like CARE. But for the later you need to train the machine which is not straightforward.
There is also a method called “Sharpening” but this just reinforces the edges in the image. Give it a try in case.
Wait denoising and deblurring are not the same
You can find some denoising method in Fiji via the search bar (try denoise). I haven’t used any so I cant really recommand one.
Yet for a light salt/pepper noise we usually just apply a slight blur (Gaussian for instance) to smooth out the noise. The blurred image is then only use to derive a mask, not to do quantification.
You can use the macro recorded first, selecting the Java language (to have a syntax compatible with Jython).
If you dont find some method you can have a look in the ImageJ API, which also offer more flexibility than the recorded commands.