I think what you're suggesting is that because we know the angle between a pair of stars, we can use that to infer the angle to the horizon by measuring the distance between each in pixels.

I don't think it's a safe assumption that there is a linear relationship between the angles. Like, two pairs of objects in the image, in which the angles subtended by each pair relative to the viewer are the same in real life, may not be separated by the same number of pixels in image space depending on the lens used.

Furthermore, this could be thrown off by even a small hill or mountain obscuring the astronomical horizon.