Trying to remove all colors except black

[jswolcott]

convert C:\temp\color.tiff -fuzz 10% -fill white -opaque black c:\temp\bw.tiff

The output file is 100% identical to the input.

I'm clearly missing some thing very simple but I haven't the foggiest.

[jswolcott]

Wow. I can not say how much of a dope I feel right now.

Thank you for pointing that out though.

wow

[jswolcott]

Alright, well I do also have imagemagick. I ran this by imagemagick (not graphics magick this time) and replaced the -opaque with a +opaque.

Now I do get changed output. However I'm getting a plain white wash. No black or any thing like it. I also tried the double %% as suggested. Not a windows expert either but we use what we are given eh?

[fmw42]

Upload your tiff image to dropbox.com and put the URL here. We need to see your input so we can test with it. Does it have any alpha transparency?

If you are getting totally white output, perhaps you fuzz value is too large. Try smaller values

What version of Imagemagick are you using? Perhaps it is too old or buggy.

[snibgo]

Code: Select all

convert C:\temp\color.tiff -fuzz 10% -fill white -opaque black c:\temp\bw.tiff

You seem to be running this from the command line. Don't double the percent sign. The rule about doubling them applies only to BAT files (scripts).

[jswolcott]

The link is on drive here:
https://drive.google.com/file/d/0B2aTfI ... sp=sharing

should be public


Oh.. Well here is relevant info. It's also outputting this error:
convert.exe: Unknown field with tag 20624 (0x5090) encountered. `TIFFReadDirectory' @ warning/tiff.c/TIFFWarnings/857.
convert.exe: Unknown field with tag 20625 (0x5091) encountered. `TIFFReadDirectory' @ warning/tiff.c/TIFFWarnings/857.

All I'm trying to do is get all the colors out of this image so that all is left is black, which is primarily text, so I can OCR it.

[jswolcott]

version mageMagick-6.9.0-Q16 windows.

[glennrp]

Try the -threshold option then, e.g., -threshold 30%. Anything darker than 30% intensity becomes black and the rest become white.

e.g.,

Code: Select all

gm convert logo: -colorspace gray -threshold 30% gmbw.png

or

Code: Select all

convert logo: -colorspace gray -threshold 30% imbw.png

[jswolcott]

Well that kind of works. However it keeps a ton of the images which are a dark color but not black... I may be barking up the wrong tree here. I'm just trying to get this simple for tesseract and find a way to automate the process.... cropping is so error prone since the images I'm taking in are not always uniform. -sigh-

[fmw42]

try this

Code: Select all

convert color3.tiff -alpha off -fuzz 35% -fill white +opaque black result.tiff

Adjust the fuzz value as desired. Note you have no pure black in the image. That is why when you use a small fuzz value it shows as pure white.

[snibgo]

"-contrast-stretch" can be useful, eg increase the contrast until 30% of pixels are black and 20% are white:

Code: Select all

convert color3.tiff -contrast-stretch 30,20% t.png

[jswolcott]

Wanted to swing by and say thanks to ya'll. I ended up using
convert color3.tiff -contrast-stretch 35,25% t.png
followed by
convert color3.tiff -alpha off -fuzz 10% -fill white +opaque black result.tiff

Worked perfectly. Removed the color backgrounds entirely and I was then able to crop and OCR the text boxes. Thanks!

[snibgo]

Your first command "convert color3.tiff -contrast-stretch 35,25% t.png" creates t.png. If you don't use t.png, it is pointless.

[jswolcott]

well yeah modified for correct file names. It's inside a script and runs on many images so.. any way. point is it worked.