Use convert for jpeg to tiff suitable for tesseract

[Paddy Landau]

I often need to use OCR (optical character recognition) to get new text onto the computer.

When I use tesseract, I get terrible results, unless I use GIMP to set the image to 1-bit monochrome (and save as TIFF).

(See "Preparing images for Tesseract" at the bottom of the Community OCR page.)

As I often get these files, I'd like to automate the conversion, rather than have to fire up GIMP and manually do it every time.

I tried using convert, but the command gives an error when I try to convert it to monochrome:

$ convert newtext.jpg -flatten -monochrome newtext.tif
convert: BitsPerSample 1 not allowed for JPEG. `JPEGSetupEncode'.

If I use convert without the -monochrome option, then tesseract puts out garbage.

Any ideas?

[kaibob]

I tried the jpg to tif conversion with the monochrome option and got the same result as you did. I was able to convert from jpg to png with the monochrome option and then to convert the png image to tif. I checked the resulting tif image and it was 1-bit and loaded fine in my image viewer. It's a bit of a kludge, but if nothing else works, you may want to give it a try.

[Paddy Landau]

kaibob, that works a treat. As you say, a bit of a kludge, but hey, it works.

Thanks.