2022-05-11: APP 2.0.0-beta 1 has been released !!!
Download links per platform:
[Solved] 1.076 Beta 3 noise added in Crop operation
I was experimenting with the new Beta and I have noticed something when doing a crop. The image that is created seems to have much more noise than the un-cropped image. I do not know if this is unique to 1.076 (I don't think so, I may have noticed it in previous versions) but it is there. If you set maximum stretch, the original is OK (in my case) but after a crop, it's very noisy in the shadows.
Here is a link to the before and after images. Set maximum stretch and look at the shadows (dark) below the horsehead.
APP is not able to inject extra noise, it would have to do that deliberately and it's not programmed to do so (extra noise can only come from using wrong calibration data and such). I see a slightly different stretch, which is done automatically in the preview as the crop now has a slightly different overall signal. If I stretch the cropped version to have approx. the same stretch (looking at the overall brightness of the background and dust), I don't see a real difference anymore (maybe now less noise in the cropped version, which shows it's the stretching which is slightly different).
Not cropped version;
Thank you for bringing this to our attention.
I have just checked, but everything is perfectly fine here. The crop operation really only crops, it does not inject noise. No pixel values are changed, if it would, it would be very bad indeed.
You might think that there is more noise because of the auto-stretching that is performed on the data with the preview filter. The autostretch manipulates 2 stretch parameters:
- the black point, B-slider
- the stretch ST-slider
If I leave the auto DPP on, this is what happens, I will zoom in with 400% (100% means each image pixel matches a screen pixel) on the shadows below the horsehead with maximum stretch of 30% BG, 2sigma, 0,0%base, first image is uncropped with max auto atretch, second image is cropped with max auto stretch, yes that does look more noisy, but is it?:
Now, if load the cropped image and apply the max auto strech, then disable the auto stretch by deselecting the auto selectbox, I will keep the stretch settings fixed for both the black point and the stretch. I now load the uncropped image and study the same field of view at 400% zoom, this is how it looks, first image is uncropped with the same stretch as on the cropped image with 30% autostrech, second image is the cropped data with the 30% autostretch. This looks completely identical :-), so nothing has changed with the pixel values by the crop operation :
Please let me know if this is clear. To study the noise, you need to deselect the autostrech. The autostrech in this case gives a different black point value on the cropped/uncropped images and is leading you to think that there is noise injected, where there in fact is none injected. All pixel values are preserved.
Ah, and I learned an even better way to show this. Thanks Mabula. 🙂
Excellent! I had somewhat suspected as much (the stretching being the issue), but I needed to hear it from the experts to confirm my expectations that indeed it was not noise.
As usual, a great product.
Thank you both for your valuable insight.