How to Remove Banding on Negative Scans?

[jtk]

Flare and heavy post-processing.

 

[138S]

and was very disappointed to find banding across many of my images. At first I thought it was dust caught in the Flexlight scanner I had been using however,

Banding happens when editing curves in an image that has 8bits per channel. Before editing curves of the image (in photoshop) go to menu Image->Mode and select 16 bits, save the images always in TIFF format, jpeg and BMG only saves 8bits per channel.

Always scan 16bits per channel, if you image is 8bits then convert to 16bits before editing the curves. You only convert to 8 bits when preparing the image that is to be distributed in the internet.

See here how banding is generated:

 

[Mainecoonmaniac]

Banding happens when editing curves in an image that has 8bits per channel. Before editing curves of the image (in photoshop) go to menu Image->Mode and select 16 bits, save the images always in TIFF format, jpeg and BMG only saves 8bits per channel.

Always scan 16bits per channel, if you image is 8bits then convert to 16bits before editing the curves. You only convert to 8 bits when preparing the image that is to be distributed in the internet.

See here how banding is generated:

I might be wrong on this, but banding is when there's a lack of data. Changing the file format will not add data to the image. The only way I know to scan high data images is to capture with more data and save in a lossless format like TIFF.

 

[138S]

I might be wrong on this, but banding is when there's a lack of data. Changing the file format will not add data to the image. The only way I know to scan high data images is to capture with more data and save in a lossless format like TIFF.

Banding happens when (for example) expanding an small range of bits to a well higher range. Say that in a 8bit image you expand a zone in the image with levels from 0 to 4 into the range from 0 to 64, then you only will have next levels after curve edited 0-16-32-48-64, obtaining 5 bands.

If the Flextight scans were delivered in 16 bits TIFF and later user saved the scans in JPG he would had an automatic conversion to 8bits per channel, if later he openned the 8bit file and edited the curves (contraction+expansion) then he could have generated the bands.

I guess that you agree that curves should be edited in 16 bits mode, and that saving images in jpg ends in a conversion to a 8bits image that later would be prone to banding...

 

[Nebbit]

Banding happens when (for example) expanding an small range of bits to a well higher range. Say that in a 8bit image you expand a zone in the image with levels from 0 to 4 into the range from 0 to 64, then you only will have next levels after curve edited 0-16-32-48-64, obtaining 5 bands.

If the Flextight scans were delivered in 16 bits TIFF and later user saved the scans in JPG he would had an automatic conversion to 8bits per channel, if later he openned the 8bit file and edited the curves (contraction+expansion) then he could have generated the bands.

I guess that you agree that curves should be edited in 16 bits mode, and that saving images in jpg ends in a conversion to a 8bits image that later would be prone to banding...

Sorry there has been some confusion here. My fault totally. My language was unclear, by banding I simply meant the bands of lines running across the image. I would have said scan lines, however since I suspect some of these lines are actually on the negatives I just opted to call them bands. My bad and thanks for the tips!

 

[Nebbit]

You could try this software on a trial basis: http://akvis.com/en/retoucher/download-image-restoration.php

The films I did were in 2008-2009 and it was all pretty much manual masking, pixel stealing and adjustment, there has been a lot of progress in defect removal algorithms...

Thanks for the tip! I'll look into it

 

[Nebbit]

This looks like dirty scanner glass if the scanner used glass.

It's a flexlight scanner so the glass shouldn't be the issue. I totally see what you're saying though. At first I thought it was dust caught in the scanner. This neg was definitely not thin which is typically when you would start seeing those scan lines. Additionally, the scanner was just serviced by Hasselblad so there should in theory be no issues. This example is from a 120 mm neg. When I went back to the lab that processed them we looked at some 4x5 negs as well where you could see that some the lines were definitely on the negatives themselves. It could be here that we are dealing with both a processing and scanner issue at the same time.