• Global community
    • Language:
      • Deutsch
      • English
      • Español
      • Français
      • Português
  • 日本語コミュニティ
    Dedicated community for Japanese speakers
  • 한국 커뮤니티
    Dedicated community for Korean speakers
Exit
0

Lightroom interpolating pixels with Camera DNG Perspective control

Participant ,
Sep 06, 2022 Sep 06, 2022

Copy link to clipboard

Copied

I have a recently updated Leica SL2 with the perspective control feature. Enabling this will insert additional data in the DNG file that corrects for both horizontal and vertical tilting.

 

With the feature enabled, importing the image into Lightroom, it displays the corrected image, and in the "Transform" tool, "Guided" is selected. So, Lightroom takes the data in the DNG and constructs 4 guide lines and then uses those guidelines to rotate the image and crops into the image to display it without borders (equiv to "Constrain Crop"). You can turn off the display of the corrected image by de-activate the "Transform" tool, and the image rotates to its original (fully framed capture). The size of the SL2 sensor is 8368 x 5584.

 

(1) Where the "Transform" tool is active (as it is on import with a perspective corrected file) Lightroom reports the image size to be 8368 x 5584.

(2) Deactivating the "Transform" tool, unrotates the image, exposes more pixels (that it had cropped with the guided lines). The image size is reported as 8368 x 5584.

 

If you export the image as a "Full Size" JPEG, both images are 8368 x 5584.

 

The bug:

In the case of (1), Lightroom must be interpolating pixels and the actual pixel dimensions of the displayed iamge MUST be smaller than 8368 x 5584. Further my "Full Size" JPEG export should result in a JPEG file with smaller dimensions than the full size of the sensor dimensions.

 

If I take an image (shot before the Perspective Control update) and apply the "Guided" corrections, I see the image rotate and I see Lightroom adjusting the Image size. Setting "Constain Crop" I lose the white borders and see a cropped, smaller image, as expected.

 

I can provide the DNG and the exported JPEGs, but wondered if anyone else has seen this problem or has any comments.

 

William

TOPICS
macOS

Views

144

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
LEGEND ,
Sep 06, 2022 Sep 06, 2022

Copy link to clipboard

Copied

Does this happen with RAW? In this case a DNG file.

 

Actually, are you sure you are using a JPEG image, as the perspective control mods in camera should have been baked into the image. The DNG instead would have the perspective control mods inside the metadata and available to LrC during Develop.

 

Also, not a Leica holder, so this all from google research. But, the perspective control in the SL2, is that not a new capability added via a recent firmware update?

https://www.reddotforum.com/content/2022/09/leica-sl2-sl2-s-firmware-adds-perspective-control/

Adobe may have not quite gotten around to supporting the SL2 perspective control, unlike the M10

 

https://helpx.adobe.com/camera-raw/using/leica-perspective-control.html

 

 

 

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Participant ,
Sep 15, 2022 Sep 15, 2022

Copy link to clipboard

Copied

Sorry, I missed your response.

 

The images I'm working with in Lightroom are DNG images. So, (with the DNG),

(1) the dimensions are "unchanged" with both the embedded "guides" for perspective correction applied, and disabled.

(2) when I export both versions to a Full Size JPEG, both JPEGs are the same dimensions.

 

The only way this can be, is that Lightroom is interpolating pixels when the "guided" perspective correction (embedded in the DNG) is being applied to the DNG. Which seems very bad to me.

 

William

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Community Expert ,
Sep 15, 2022 Sep 15, 2022

Copy link to clipboard

Copied

I do understand that you may not want unrequested "scaling" interpolation to happen, if that is avoidable. What you say about Constrain Crop behaviour makes sense and it seems odd that the two cases act differently. As commented already, proper support for this may not yet have been quite resolved. Or perhaps it has been done this way for consistency with what this feature does in camera JPGs?

 

But that said, to put this in proportion, a perspective transform must involve lots of pixel interpolating in any case: picture content is getting squeezed, stretched, twisted in complex varying ways and amounts all over the image. Besides any transforms by profiled lens geometry corrections (if enabled). Besides any Upright mode, manual transforms or Crop levelling that may also be set.

 

AFAIK the only way to preserve the original sensor-demosaiced pixels without any interpolation would be to disable lens profile corrections; and to forego all other kinds of calculated transform too, including preventing the camera's guided corrections from being implemented.

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Participant ,
Sep 15, 2022 Sep 15, 2022

Copy link to clipboard

Copied

LATEST

Thanks for the clarification, makes sense and hadn't thought it through to that extent.

 

I think the difference between the "Constrain Crop" and the in-camera application is what raised the issue for me, and these should be consistent I think.

- I can do a very exagerated tilt in the camera, with say 1/4 or more of the visible pixels "cropped out" and LR will still report the image size as the full-sized sensor dimensions. That seems very inconsistent to me and introducing alot of potential for scaling artefacts that I'd rather have a more direct control over if I need that many pixels.

 

Thanks!


William

 

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines