jmbhughes / noise_gating

reducing shot noise using regional fourier transforms

Hi, I'm Marcus.

• I work at the Southwest Research Institute in Boulder, CO.
• One of my main priorities is developing the Science Operations Center data reduction pipeline for PUNCH, an upcoming NASA mission to study the solar environment. Watch our code get released!
• I also work on the CubIXSS mission.
• I'm interested in machine learning applications in solar physics, with specific experience in thematic maps. I'm working on an improved version called CORD Maps.
• I'm building Barnsley, a Rust app to explore iterated function systems,with a web-based GUI.

noise_gating's Issues

Matrix size mismatch: image_section and hanning_window

Hi
Just stumbled on your toolbox looking for DeForest filter, really great job. I'm looking to use this filter on brain imaging data.

I've made it most of the way through your routines (removed arguments for now), except that the hanning filter and the image sections don't seem to be matching in size (see below). I may have a weird image stack shape, but it should otherwise work. I added +1 to the section function that slices from self.image_cube and am running it now, but I didn't think that's what you intended.

Any advice on why these are not matching?

---

cat@catlab:~/code/noise_gating$ python noise_gater_3d.py
Making image cube
(45, 80, 63500) #My image stack shape
Cleaning image, this may take a while!
10 35 3 #xstart, xend, xstep
10 70 3 #ystart, yend, ystep
10 63490 3 #tstart, tend, tstep
3808800 10000
Traceback (most recent call last):
File "noise_gater_3d.py", line 212, in
clean_cube = ng.clean()
File "noise_gater_3d.py", line 165, in clean
t-self.twidth//2 : t+self.twidth//2] += section
ValueError: operands could not be broadcast together with shapes (4,4,4) (5,5,5) (4,4,4)

EDIT: Ok, I see now that there's an integer divide which rounds off and can make the hanning_window different than the cube being processed. So I have to choose an even step as well as probably a larger FOV.