I use GPUImage to convert image to Black&White, and then I use TesseractOCRiOS to scan text on that image, but output is not good result. So I want to detect text on image and save to another image with text only to prepare for OCR to have good result.
Asked
Active
Viewed 1,115 times
1
user1916184
- 135
- 2
- 14
-
You can find text with Opencv [Scene Text Detection](http://docs.opencv.org/3.0-beta/modules/text/doc/erfilter.html) and the crop the subimages with text. – Miki Aug 04 '15 at 11:32
-
Thanks @Miki, but I don't know the way to write it in `Objective-C`. Could you write a demo small app? – user1916184 Aug 05 '15 at 02:24
-
GPUImage doesn't have a simple method which you can use to detect text. That's a very complex problem. Which methods did you try? Several methods are described here: http://stackoverflow.com/questions/10206526/how-to-detect-text-area-from-image If you plan to use OpenCV - you can use it with C/C++ only. – Cerovec Aug 08 '15 at 21:17