Is there a general rule of thumb for when it is better to include a bit of border when converting? I am DSLR scanning.
I mostly convert with a border buffer size that spares me from cropping each take individually. In a few cases, converting a cropped image without border buffer yields images that are closer to what I want them to be.
Rule? Find your way and stay open for an occasional detour.
I leave the border/flim edge in the frame when copying. I use that for initial NLP white balance, then crop it out in the final conversion. I’m using a 45 megapixel camera, so I don’t mind losing a few pixels in the crop.
Thx. I understand that part. There are simply some images that convert worse without cropping out the border, so I’m wondering if there is a way to predict which will convert better with the border uncropped. I don’t always have the time/patience for trial and error, and a simple heuristic in this case would help.
Of course staying open for a simple detour is a good rule. What I’m wondering is if there is a way to predict those few instances the border should be left in before I convert a roll. I don’t always have the time/patience for trial and error.
It’s not trial and error. Each roll will have a slightly or greatly different base (orange) color due to film type and processing. For that reason, I white balance each roll individually. Based on my experience, that should always be your starting point.
If there are color issues after white balancing on the film edge, that’s a completely different issue. It can be due to poor film, exposure, processing or age of the negatives. When I see those issues, I deal with them as much as possible in Lightroom Classic. I always have LrC create tiffs so I can work the sliders normally in my workflow. Once I do the white balance on the first image in a roll, I crop that image and then sync the white balance and crop in Lightroom Classic to the rest of the shots on that roll.
Haven’t managed so far. I follow my feelings, and I occasionally have to do a second run. Alternatively, you could work with virtual copies and select the conversions you like best.
My experience so far is that in scenes with very limited colour variation, the conversion including some border is usually better. Just recently had some shots taken during a flight (lots of blue sky and some clouds) and the pre-cropped conversion looked really strange.
Now I usually start leaving some border when I know in advance the image is quite monotone. Admittedly usually I still do a virtual copy check without boarder to be sure I am better off 
This is tracking with what I’ve experienced so far, but haven’t quite been able to put my finger on… the images I converted that made me wonder were somewhat monotone cloudy beach photos that were converting very strangely with the normal process (white balance edge and crop or buffer border out before converting) but looked much better with a bit of border left in.
And yes, virtual copies are always the way to go.
I think I read somewhere that images without a good black point in them should have some border left in, as well.
IMHO, the borders should be left out of the image-to-be-converted all the time. Once you sample the borders for white balance , exclude them completely from conversion. After all the whole science of NLP is to adjust and properly map image histograms for each RGB channel. Inclusion of border simply skewers histograms as non image data suddenly goes into play. HTH
1 Like
I cannot agree with you there, it certainly works for 99% of my images but not for all. Here is an example comparison of a shot taken during a flight as mentioned above. I sampled the border for white balance before conversion, left one included some border, right completely cropped. In NLP no white balance and all default settings. One can start tweaking of course until it looks reasonable, but it’s a poor baseline to start from.
Let’s just agree that NLP cannot “always” handle all sorts of shots. It handles very well shots which have mostly full range of tones and colors like most landscape or cityscapes have. For outliers like your shot from a plane the standard algorithm does not work as the shot have very strong predominant color and low contrast. Including border in that case is just a hack. If that works - fine, it saves you the trip to Photoshop. My point is - having non image data in conversion is exception from the principles upon which NLP works. See my article Negative Lab Pro by Vlads Test Target | On Film Scanning for details. But again if some hack works - fine, but that is not a new rule or recommendation for folks to deliberately include borders in conversion. That’s just my 2 cents.
Sure. Makes sense. But also, that was the point of my post… asking about identifying exceptions to the rule before converting.
1 Like
Checking the histogram can help.
- If the histogram is narrow and on the left side, borders might help.
- If a capture is almost monochrromatic, NLP will probably strunggle, whether borders will help should be discovered case by case…and roll analysis should be tested and might be beneficial, mostly if other captures have “richer” histograms.
1 Like