Plotting RGB(Multi channel) image using ants.plot()

[J-Yash]

I am trying to read an rgb image(.jpg) and plot it using ants.plot() but it does not plot the image properly and instead plots noise. Here is my image : (This is just a sample image I am using to test plotting of rgb images)

Code to read/plot image : im = ants.image_read(PATH_TO_IMAGE) im.plot()

The image that is being plotted looks like this:

The ANTsImage object header using print(im) gives:

Is there a specific way to use rgb images in ANTs? Please advise.

Thank you.

[ntustison]

I stumbled across this the other day and had to use imageio as a workaround. Note that if you save starry_night_ants as a nii.gz file, you can then read it back in using ants.image_read. I have this on my list of things to fix but just haven't gotten to it yet.

[J-Yash]

1. So I tried using your suggested method to get the ANTs image but plotting that using ants.plot() still plots noise. Is there another way I could plot this ANTs image?

2. Even though the plotting is buggy, does the ants.registration() method work fine on such RGB images?

[ntustison]

It works for me.

>>> starry_night = ants.image_read("starry_night.nii.gz")
>>> ants.plot(starry_night.vector_to_rgb())

Even though the plotting is buggy, does the ants.registration() method work fine on such RGB images?

Multi-component images, including RGB, have to be split up and any components beyond the first would need to be plugged into the multivariate_extras argument of ants.registration.

[J-Yash]

That solves my issue. Thank you. Although I was curious: Does using an RGB image lead to better registration? Or simply converting my image to grayscale(Or simply using one channel) and then doing the registration would give similar results?

[ntustison]

Medical imaging doesn't typically utilize RGB images so we don't have direct experience. However, since the RGB channels are highly correlated and since our go to metric is cross- correlation, you probably won't gain much using all three channels separately.

[J-Yash]

That's very helpful. Thank you.