snibgo wrote: ↑2018-05-02T13:53:12-07:00
For the general case, if you have many posters to process, algorithms exist for detecting text. They detect edges that are roughly parallel to each other, within a certain range of distances. The results are readable by OCR. I don't know if those algorithms have been implemented in generally-available software.
snigbo your advice is as deep as it is prescient. I'm not joking when I say you are really really
really smart! I'm not only taking a screenshot of your comments, I'm ripping the entire thread for a permanent reference! You explore alternatives I just never even thought of. Wow. Thank you.
snibgo wrote: ↑2018-05-02T13:53:12-07:00
(cont'd) For movie posters, that's probably overkill, because finding the text isn't difficult: it's a rectangle near the bottom that contains high-spatial-frequency data. Crop to that rectangle, and whack up the contrast. Assuming light letters on dark background, we make 50% of pixels black and 10% white:
Code: Select all
magick im-rev-999999.png -crop 1684x235+84+72 -channel RGB -contrast-stretch 50,10% +channel x.png
I referenced my flatbed scanner in my lead post, but it just so happens I
also have OmniPage 15 (that's a commercial
OCR program for anyone just learning about some of these terms). I never thought of tackling my little poster projects that way. It's a curious approach. I don't doubt OmniPage can recognize the text, but what you're asking is (if I'm reading your comments correctly) would it produce the
font.
For intellectual property & trademark reasons OmniPage can produce a digital-file/clipboard-copy of plain text -or- as close a
typeset replica of the original font if that font was already installed on your computer.
If your project were for example, the lengthy chapter of a book with title headers and footnotes, this might make sense. But a roll of cast members, titles, studio, runtime, release year, address data etc. amounting to the equivalent of a large paragraph? Follow the thought to its conclusion:
You might as well just type it.
Yes, I have attempted
even that, to my shame!
(it could be worse; my hobby could be collecting Model T Fords!).
So
snigbo I'd like to return to the most recent of your scripts. Please note:
- I've changed the name of file we're operating on from im-rev-999999.png to GQ-99.png
- The batch file should be named GQ-99.bat
- Both files have been copied directly into my ImageMagick program folder
The environment is:
Z:\WINDOWS\system32\cmd.exe
Z:\[PROGRAMS] XP\IMAGEMAGICK\GQ-99.png
Z:\[PROGRAMS] XP\IMAGEMAGICK\GQ-99.bat.
The batch file reads:
Z:\[PROGRAMS] XP\IMAGEMAGICK\convert magick GQ-99.png -crop 1684x235+84+72 -channel RGB -contrast-stretch 50,10% +channel x.png
and the error I get is
Code: Select all
Windows cannot find 'Z:\[PROGRAMS XP]\IMAGEMAGICK\convert magick GQ-99.png -crop 1684x235+84+72 -channel RGB -contrast-stretch 50,10% +channel x.png'. Make sure you typed the name correctly, and then try again. (SNIP)
whether I manually open a command prompt -or- perform a "Run As" Administrator on
GQ-99.bat
I've Googled for a Beginner's Tutorial on ImageMagick but the content on the imagemagick-org site has been removed, and the one PDF I was able to find is focused on the "Magick++ C++ graphics library". If I haven't lost you yet
snigbo I would very much appreciate active links to Beginner's guides and tutorials. Thank you everyone. I apologize for the length but I am a Writer
(no, we'd never guess!)