My first thought was "It's winking", too.
As an aside, the two images of Betelgeuse are really good examples of an issue I've struggled with a lot: controlling the information content in a visualization image.
We do not actually know what the true resolution of the imaging machinery are with relation to the image dimensions. The image we see could very well be reconstructed from data from as few as 25 separate imaging sensors ("pixels"), but smoothed out and "prettified" (filtered with a low-pass filter) so it doesn't just look like 25 squares of slightly different colors. There is a huge risk of accidentally doing a computer-assisted
Face on Mars; but instead of light and shadow playing a trick on the viewer, the added interpolation/smoothing/filtering done to the digital data does it.
(In my case, the problem is that when visualizing molecules or atomic systems, displaying them as shiny spheres with perhaps sticks between them to illustrate bonds, gives completely wrong intuitive picture. Atoms do not have clear boundaries at that scale, and when bonded, most definitely do not stay spherical. So, using "less fidelity", or just non-photorealistic renderings like something from the graphics artists' toolbox, one can avoid causing such mis-conceptions and mis-intuitions by controlling the information conveyed [by adding non-realism!]. At the atomic level, especially when you have lattices (perhaps a metal), small clusters, and individual atoms, just defining exactly what constitutes
a collision and how energy or momentum is exchanged, is
hard. They don't bounce off each other like people in a mosh pit, or pool balls, but tend to do pirouettes around each other like classical dancers.)