Astro DIY: Sun Photos

Testing my new AVX mount. With a level bubble, compass and GPS coordinates, I was able to get it aligned fairly accurately, enough to capture the sun and have it remain in the field of view on the order of minutes.

The setup. A 127mm Maksutov-Cass on a Celestron AVX mount, with a Baden solar filter. The camera used is a DSLR Canon Rebel T3.

When I took image, I was a little disappointed, there were some defects (which at first I thought were sunspots :-( ). This is due to dust motes.

Sun. You can see defects in image

Unfortuntaley, I didn't take a flat field. However, I did have a few images of the sun that were shifted, since my tracking wasn't perfect. So I masked out the dust motes:

Masking defects

And chose two images decently shifted apart, and re-shifted them. There are sophisticated methods out there for this but for this application, I decided just to shift manually and look at the absolute difference of the image and the second image reshifted. When they align, you should see a uniform noisy sun. (Else, you'll see some bright regions which represent the overlap)

Taking two images, top left and top right are two images where the sun moved in field of view. If I shift them back by the correct amount and take difference, we see a nice noisy image as lower left figure.

After creating a mask and knowing the shift, you can then write some quick code to average the two together, as below:

def combinephotos(imgs, shifts, mask):
    imgresults = np.zeros_like(imgs[0],dtype=float)
    imgcts = np.zeros_like(imgs[0],dtype=float)

    for i in range(len(imgs)):
        imgresults += np.roll(np.roll(imgs[i]*mask,shifts[i][0], axis=0), shifts[i][1],
                axis=1)
        imgcts += np.roll(np.roll(mask,shifts[i][0], axis=0), shifts[i][1],
                axis=1)

    imgresults /= imgcts

    # parts that average more than one image, add poisson noise
    #w = np.where(imgcts > 1)
    #imgresults[w] = np.random.poisson(imgresults[w])
    return imgresults, imgcts

shifts_array = [
        [0,0],
        [-28, -51],
        ]

img_final, imgcts = combinephotos([Grey1, Grey2], shifts_array, mask)

There are much more efficient methods of course, but this allows more control over your data, and better understand exactly what it is you want. For example, I don't like inpainting. This method does not require any inpainting or blurring whatsoever. This is the raw data. The only potential issue is the loss of pixel resolution from the error in the estimation of your shifts. However, there are ways to actually get subpixel resolution, something I'll ignore here because this image is already quite large!