Can I make OCR to use threshold? Topic is solved

[liij]

On the site http://capture2text.sourceforge.net/#download there is command like this

Code: Select all

  --tess-config-file <file>          (Advanced) Path to Tesseract configuration file.

My script works, but usually OCR does not recognize number correctly. I know I can fix this by changing threshold, because numbers on the screen are almost white compare to the background.
I am searching for ''tesseract'' command to change threshold

Code: Select all

q::
TempFile := A_Temp . "\temp_capture.txt"
FileDelete, %TempFile%
RunWait, C:\Users\xxxx\Desktop\Capture2Text_CLI.exe -s "1805 22 1897 33" -d --whitelist 1234056789/ -o %TempFile%,,Hide
sleep 1
FileRead, CaptureText, %TempFile%
FileDelete, %TempFile%
var := StrSplit(CaptureText,"/")
MsgBox % var[1] "`n" var[2]
MsgBox %CaptureText%
return

This outputs 152/188 according to script

debug_capture.png


debug_enhanced.png


I was trying imagemagick with many commands and thresholds, but it does not work correctly on windows7 in my opinion

Code: Select all

"C:\Program Files\ImageMagick-7.0.7-Q16\magick.exe"     convert -density 150 -threshold 60% rose.jpg out.jpg

There is image barely visible below

[JoeWinograd]

Try this ImageMagick command on your captured image:

Code: Select all

magick.exe convert debug_capture.png -geometry 500% -threshold 90% white_on_black.png

That creates a white-on-black image. The OCR packages that I tried did better with a black-on-white image, which you can create with this ImageMagick command:

Code: Select all

magick.exe convert white_on_black.png -colorspace HSI -channel B -level 100,0% +channel -colorspace sRGB black_on_white.png

That created this image:



Capture2Text 3.9 OCRs that perfectly (does not with the white-on-black.png file):



Regards, Joe

[liij]

@JoeWinograd Great job with commands, but ImageMagick does not work on my pc correctly. This is how I am running this ( i know I can run wait .bat with ahk).
I made geometrythreshold1.bat from notepad with code magick.exe convert debug_capture.png -geometry 500% -threshold 90% white_on_black.png. This is what I got:

After that I made colorspacechannel.bat with code magick.exe convert white_on_black.png -colorspace HSI -channel B -level 100,0% +channel -colorspace sRGB black_on_white.png

Definitely I am doing something wrong. My system is windows 7.

[JoeWinograd]

I executed the commands in a Prompt window. Since you put it in a batch file, you need to escape the percent signs, which you do with a double percent. So in a batch file:

Code: Select all

magick.exe convert debug_capture.png -geometry 500%% -threshold 90%% white_on_black.png

magick.exe convert white_on_black.png -colorspace HSI -channel B -level 100,0%% +channel -colorspace sRGB black_on_white.png

Btw, I'm on W7 Pro 64-bit running ImageMagick 7.0.7-3-portable-Q16-x86. Regards, Joe