There are usually small variations between images when you shoot a stack for HDR, especially shooting handheld. Wind may cause larger variations, as well as other moving entities within a shot (such as a pedestrian in the distance).
Photo editing software is good at dealing with these variations. However, certain items, especially complex/movement-prone ones like tree branches, may require manual intervention to avoid "ghosting".
Shooting
We want to give the software the best source material we can.
- Shoot on a tripod; if handheld, be sure that your shutter speed is fast enough that your movement won't cause blur, even on the longest exposure.
- Avoid long delays between exposures (shoot the stack as fast as possible).
- Shoot plenty of images for the stack, but try combining only some of them for the best result. The more images, the more variations the software must deal with.
- Try to use settings which give at least one decent exposure at a fast enough shutter speed to eliminate the movements you can't control (like wind). Other images in the stack may have movement.
Deghosting
"Ghosts" are removed by deghosting.
I don't use Aurora often, but this tutorial discusses the deghosting process in Aurora and the results look good: https://youtu.be/7OuIfowv8U8
Lightroom supports automatic deghosting. Photomatix (my preferred tool, and similar to Aurora) supports deghosting and does a fantastic job of it when starting from good source material.
You can also manually recreate and/or tweak the repair using a tool like Photoshop.
- Find an image which is clear.
- Copy the portion of the clear image over any blurry portions of the composite.
- Adjust the copied "piece" so that it blends with the composite image.
Examples
This image was shot during heavy wind and rain. The tree above the truck showed movement, so I selected it for deghosting in Photomatix.

The flag in this image showed movement. I recall manually editing this image using the aforementioned technique to blend a sharp version of the flag into the composite image.
