As of Recoll 1.43.3, the alternate Python rclimg.py and rclimgp.py handlers can execute OCR on image files.
The default image handler is the Perl/exiftool-based rclimg script which has not been extended for running OCR, but it is quite better for extracting metadata tags.
The difference between the two alternate handlers is that rclimg.py uses the Python exiv2 interface and rclimgp.py actually executes the Perl handler for extracting tags. On Windows only the latter is supported.
Especially on Windows, using rclimgp.py makes image tag processing much slower. This is not significant when actually performing OCR, which takes even much more time, but it means that it is better, if possible, to use a separate index for image OCR (there is no way to use different handlers for different areas of the file system within a single index).
For performing image OCR, you need to tell Recoll to use the alternate handler and
also to enable OCR by setting the imgocr variable. This is done by
editing two text files in the index configuration directory (by default
~/.recoll under Unix-like systems or MacOS systems, or
C:/users/[you]/AppData/Local/Recoll under Windows.
Example configuration for Unix-like systems or MacOS systems:
In mimeconf
(e.g. ~/.recoll/mimeconf):
[index] image/gif = execm rclimg.py image/jp2 = execm rclimg.py image/jpeg = execm rclimg.py image/png = execm rclimg.py image/tiff = execm rclimg.py image/x-nikon-nef = execm rclimg.py image/x-xcf = execm rclimg.py
Of course you can also only use a subset of the image types, and the rclimgp.py handler could be used instead.
On Windows you must use rclimgp.py (because there is no Recoll support for the exiv2 library):
[index] image/gif = execm rclimgp.py image/jp2 = execm rclimgp.py image/jpeg = execm rclimgp.py image/png = execm rclimgp.py image/tiff = execm rclimgp.py image/x-nikon-nef = execm rclimgp.py image/x-xcf = execm rclimgp.py
In recoll.conf, by example for Windows:
ocrprogs = tesseract tesseractcmd = C:/Program Files(x86)/Tesseract-OCR/tesseract.exe tesseractlang = eng [c:/path/to/my/images/folder] imgocr = 1
the tesseractcmd setting is usually not needed on other platforms (the
command will be in the PATH)
