Hi, i was wondering why I keep getting small pockets of noise in my images when bracketing and then merging into a HDR image. I am taking 5 shots, 2 stops apart.

Deghosting does indeed cause this. If an area is deghosted, then the pixels of only one image are used in that area, because using two or more images creates that ghost. Because that is usually a darker image, you will get more noise in the deghosting area.

If you really must deghost, and the noise is bad and denoising the HDR does not work, then what you could consider is this: use Denoise first on the brackets, then use the new Render to DNG (because that flattens the denoised pixels into the file) and finally use HDR merge on those DNG files.

Thanks Conrad. No it was all done on M setting and iso was set at 400. Upon re-stacking without the darkest image it cleaned up. Also for some reason when I re-stacked the HDR images, but this time put the de ghosting level to “none” it also disappeared. 

I think I’ve had this happen once or twice until I looked at the images more closely.

Do the areas of noise correlate to a tonal range, like the darkest image in the bracket? If so, the problem might have something to do with how the camera was set to bracket the shots…how was that done?

For example, if the camera was set to auto-bracket, and also set to auto ISO (ISO speed is allowed to shift up to help expose a darker area), then if one exposure has a significantly higher ISO speed than the others, that source image may end up noisier than the rest. The differences in noise level between exposures then becomes obvious when merged.

If you look at the EXIF metadata for the source exposures and you think that happened, if you have a chance to reshoot it you could lock down the ISO speed to a setting that isn’t noisy, then the next time you bracket, change only the shutter speed. (Changing aperture may cause exposures to have different depths of field, which could also look inconsistent when merged.) If you’ve having the camera auto-bracket, that typically means do it in aperture priority mode with ISO speed manually set.

If you confirm that, when viewed separately, one or more exposures is definitely a lot noisier than the rest, but you don’t have a chance to reshoot, there are still some things to try…

One potential solution is to see if the exposure that had the obvious noise can simply be left out of the HDR merge. Some camera sensors have enough dynamic range that you don’t need 5 images to get good data throughout the target tonal range. With some newer cameras you might need only 2 or 3 images, depending on the dynamic range of the scene.

Another option if you can’t reshoot is to try running Denoise on the noisiest source exposure(s), then merge, and see if that clears up the post-merge result.

If your camera sensor is known for excellent dynamic range, you could try picking one of the middle exposures alone and apply the Adaptive Color profile. If that exposure has no blown highlights or clipped shadows, Adaptive Color could get you to a good highlights/shadows starting point with no need for HDR merging. I am finding more and more than some of my brackets aren’t needed because if the sensor is good enough to capture the entire tonal range of the scene, Adaptive Color does a good job of optimizing for highlights and shadows from a single good exposure. But Adaptive Color works well on HDR brackets too.

I understand this a not a great photo but you can see the  noise on the HDR image, but is not in any of the 5 bracketed shots