Milky Way, Big Meadows, Skyline Drive, Shenandoah National Park,April 17, 2021. 2021年4月17日，银河，大草地，香浓国家公园。 (V)
I'm trying to capture the full band of the Milky Way in two different experiments. With the first experiment each part of the scene is captured by a single image. With the second each part is captured by a series of 12 images which are stacked together in post-processing to increase the signal to noise ratio. The challenge with the second experiment is that it takes about an hour to cover the entire scene due to the 12x number of images required (the first only needs several minutes) and the Earth rotates quite a bit over that period, so the relative position between the stars and features on the ground will have changed drastically, making it more difficult to stitch the scene together.
I tried both Hugin (free) and PTGui (trial version) for stitching. PTGui felt much more superior at least with images from experiment 1. It needed less user adjustment to the control points generated by the software, did better in both matching features and blending, and was lightning fast. However, none of the images shown here was processed by PTGui because the trial version adds watermarks to the results. The PTGui matching of images from experiment 2 right out of the gate was bad as well, as expected. I didn't spend time tweaking it since the result would be watermarked anyway. Instead I spent hours playing with Hugin (all manual control points and adding masks), and focused on aligning the Milky Way (only top row of the photos were used) once it proved to be difficult to align the foreground as well.
Update (9/10/2021): With a licensed copy of PTGui I tried both stiching again. It was easy to achieve good results with images from experiment 1. I spent significant amount of effort on experiment 2 images and the result was just not usable due to misalignment. So similar to the Hugin case, I then focused on the sky/star only and it was again much easier. The result was so good that I stitched the ground rows separately into another panorama, and then tried to stitch the two panoramas together. There was a bit of learning curve involved with using panoramas as input and the resulting alignment was far from perfect, however, it was close enough that they can be merged in Photoshop, taking advantage of the large shadow areas below the tree line where not much detail was visible. In the end the result was pretty good. These two images were added to pages 1 and 2 of this album.
