Remove Vignetting from thousands of images

Question

I am doing historical research for which the only document source is microfilm typically created in the 1950s by technicians who did not understand the characteristics of this then new technology. Flat-bed scanning was not available at the time and even flourescent illumination was rare so the resulting images suffer from vignetting both as a result of the camera system, which used an inappropriate f-stop, and the illumination from incandescent spot-lights. This is further aggravated by the most common implementations of microfilm viewers in libraries, which use a lens to focus an image on a camera sensor instead of scanning the microfilm. The resultant image files thus have several layers of vignetting and optical distortion. And there are, of course literally millions of reels of microfilm, each with typically over 1000 images. The Family History Library of the Church of Latter Day Saints has fortunately done flat-scans of the several hundred million microfilm reels they have buried under a mountain in Utah, but that only applies to those documents which

1. are of interest to the LDS objective of ensuring that all ancestors have been baptized, and
2. whose original owner has permitted the FHL to make available on the website FamilySearch.org.

For my research I need access to thousands of pages of original documents for which those conditions do not apply. All of the suggestions about how to remove original vignetting from such images that I can find here or on other discussion boards are simply too much work to apply to the thousands of images. Since vignetting is an intrinsic problem with any photographic image I do not understand why nobody has developed a filter which models the actual mathematical properties of vignetting.

I am also extremely limited in funding.

Answer 1

If we talk about about digital images and the vignetting is more or less constant you can use for example XnView MP, select images, create batch convert and add as action Vignetting. Then select appropriate values and run the batch (the values on the image are extreme to demonstrate the effect):

P.S. No affiliate with the developer(s) of XnView

Answer 2

First things first: If the vignette is too dark, then recovering data from scans is next to impossible(without severe loss of quality). If the vignetting is light, then data can easily be recovered, but the image will appear washed out unless carefully toned.

In any case, here are the solutions that came to my mind.

If you think you can finish them all in 30 days, get yourself a lightroom trial version. Apply the vignette correction you want to the first image,then select all images and sync settings and export them back out. DONE!!!

else,

You could also simply get someone with a little programming knowledge to help you out. It is quite easy to write a python script which uses OpenCV or ImageMagick. (IMHO Open CV programming is more complicated than ImageMagick but installation of ImageMagick is not as easy as OpenCV).

If you want an easy approach to ImageMagick using python, I'd suggest installing wand (pip install Wand). Before Installing Wand, you need to have ImageMagick Libraries installed.

Follow the steps in :

https://docs.wand-py.org/en/0.3.5/guide/install.html#install-imagemagick-windows

download either ImageMagick-6.9.12-19-Q8-x86-dll.exe or ImageMagick-6.9.12-19-Q8-x64-dll.exe

depending on the python runtime you have.

download from https://download.imagemagick.org/ImageMagick/download/binaries/

if you face any errors after that, refer to: https://docs.wand-py.org/en/0.6.6/guide/install.html#install-imagemagick-on-windows

In order to get started with the vignette removal, refer to https://www.geeksforgeeks.org/wand-vignette-method-python/

The above link is only to provide the skeleton of your programme. You need to modify the code to suit your needs later on. For example, create a variable to measure the resolution of your image and then, using your judgement, pass arguments into the vignette function as needed(The start and end radius will be fractions of the resolution {the vignette intensity and feather radius depends on source images}).

Fine Tune the parameters as you see fit and add a few lines to recursively search through the folders in which you have images saved.