Hi,]i am using subtitle edit to ocr the text images, that was created from using videosubfinder, a nice free program
But , i found that parts of video parts of the original video, are captured as txts and hence , could not correctly translate to required, say, english language.
it is ofcurse capturing the letters correctly.
is there anyway of this kind of images from being caputured, other than manually going and deleting all those wrong texts images.
+ Reply to Thread
Results 1 to 7 of 7
Are you saying that non-subtitle text such as street signs are being captured by the program?
You will probably have to clean that up manually.
Hi, what i meant was ,
we do crop the portion at the bottom of the vsf when wew scan the txt images .
When it is doing so, some video parts resembling the texts are also captured, leading to the non recgonition at the conversion of txt images to the text with ocr tesseract by subtitle edit program.
There these characters are captured as separate images, which naturally could not be recognized by any tesseract.
This is i am asking , is there any way to filter those images at the stage of creation of images itself.
Hope that this helps you to answer
Instead of some cryptic description why don't you just post a sample image that shows the problem?
But if I'm understanding you, unless there's something about the extra text that's makes it easily distinguishable from the subtitles (like a different color, very different brightness, etc.) you won't be able to remove them.
Last edited by jagabo; 27th Jun 2020 at 20:23.
hi, one sample video parts looks like text has been captured
I have no idea what jraju is trying to do. However, if you want to remove something that is in the video itself, then "Delogo" is one option. Here's a short tutorial I created many years ago. It shows how to remove a simple spot from sensor dust, but it works for date/time stamps on home video; "bugs" that TV networks embed on their broadcasts; and subtitles.
Last edited by jagabo; 27th Jun 2020 at 22:27.