Getting the initial image file from the scanner is very simple, and should work with this command:

  scanimage --output-file='photo.pnm'

Now, every scanner has various options that you can pass via scanimage. For instance, mine has the option --scan-intent Photo, but I couldn't notice any difference when using it. Therefore, I recommend you just use the basic command above, but if you're curious, run scanimage -A to see all the option available.

The output format is PNM, which is a raw, uncompressed image format. It's also fine to start with lossless formats like PNG. Once he are happy with the result, we will convert the image to JPEG.

This is the resulting image fresh from the scanner (I chose a Napoleon 0€ souvenir banknote as the sample for this post):

You can see that the result of the scan has the size of an A4 sheet. What is not as visible is that the scanner added a white border around the resulting image. You can see it if you focus on the upper left corner. In any case, we have to remove all the empty space. This is done like so:

  magick photo.pnm -fuzz 15% -trim photo-trim.pnm

The "fuzz" percentage above can be tweaked according to the results you get. The more noisy or grainy the initial image is (e.g. due to dust or imperfections on the scanner bed or background), the bigger this percentage has to be. In my case, 15% worked fine, but 10% would only trim part of the white space.¹

After the fuzzy trim, the image might look good already. This is true for our example above. However, in some cases (especially with darker photos) we are left with very small but still visible white edges. This happens because the scanned item can never be placed perfectly square on the bed of the scanner. Furthermore, the scanner itself might have a slight misalignment, or the item might move when lowering the lid. You can see this in the following example:

The quick and easy option here is to use the shave operation, like so:

  magick photo-trim.pnm -shave 10x5 photo-trim-shave.pnm

This example command will "shave" off 10 pixels from the left and right edges, and 5 pixels from the top and bottom edges. In the unlikely scenario where you want to remove a different amount from each edge, things are a little more complicated. In the example below, one trim and four chop operations are combined to achieve a fine-grained control over the final result.

  magick photo.pnm -fuzz 15% -trim +repage \
      -gravity East -chop 5x0 \
      -gravity South -chop 0x8 \
      -gravity West -chop 4x0 \
      -gravity North -chop 0x15 \
       photo-trim-chop.pnm

The last thing we need to do is to resize and format the image according to the intended use. For instance, my blog currently has its content laid in a column 768 pixels wide, so it makes sense to use this size, unless I want to leave my readers the option to open the image in its full size. For archival purposes, I would keep the files closer to their original size so as to not lose too much accuracy and detail. As for the format, JPEG offers great compression-to-quality ratios according to your needs:

  magick photo-trim-shave.pnm -resize 768 \
      -quality 80 -interlace JPEG -format jpg \
      photo-trim-shave.jpg

The number passed to the -quality option is a percentage, so 100 is the highest quality, but also the biggest file size. You can try different values and choose the best one depending on what compromises are you willing to make in order to have a smaller file size.

To summarize, here is a complex magick command that will perform all the processing steps we described, all at once.

  magick photo.pnm \
         -fuzz 15% -trim +repage \
         -shave 10x10 \
         -resize 768 \
         -quality 80 -interlace JPEG -format jpg \
         photo.jpg

To help you with scanning multiple photos or documents, scanimage offers very useful options:

  scanimage --batch='photo-%d.pnm' --batch-prompt --source Flatbed

This command will produce files named photo-1.pnm, photo-2.pnm, and so on, for as many scans you make. Before each scan, you can take your time to position the photo on the glass bed, then lower the lid and press ENTER to start the scan. When you have finished all the photos, press Ctrl+D to signal the end of the batch. You might not need the --source Flatbed argument if you only have a basic scanner. My scanner also has an automatic document feeder, which is the default source for batch scanning, and this is why I need to specify the flatbed explicitly.

I suggest that you don't mix in a batch photos with different characteristics. If some images will differ too much from the rest, they might require special processing. For instance, I have some photos printed on paper that is bigger than the photo itself, and so they have white margins that have to be removed. These should not be mixed with photos that occupy the whole paper sheet.

Issuing the same command over a series of files is a rather general problem, so I won't go into details here. Essentially, we ruse the same commands as presented in Part I while making use of the shell's for loops, which looks like this:

  # Inline (compact) "for" loop
  for f in *.pnm; do magick "$f" -quality 90 "${f%.pnm}.jpg"; done

  # Multiline "for" loop
  for f in photo-*.pnm; do \
    magick $f -fuzz 15% -trim +repage \
      -gravity East -chop 10x0 \
      -gravity South -chop 0x10 \
      -gravity West -chop 10x0 \
      -gravity North -chop 0x10 \
      trimmed/$f
  done

In the first example, we operate on every file ending in .pnm in the current directory, and we put the output in a file of the same name, but with its extension changed to .jpg (this is what the ${f%.pnm}.jpg does).²

The second example is what we might use for the files scanned in Step 1. The photo-*.pnm pattern will only match files like photo-1.pnm or photo-123-abc.pnm, so you reduce the chances of accidentally editing wrong files. Note that the output files are created in a subfolder called "trimmed" — useful to avoid cluttering the main subfolder.