-
Technology to recognize and synchronize the text of ClassVideo with slides.
-
There are two methods for text recognition (OCR).
-
The geometry-based method is less susceptible to noise and is easy to implement with good efficiency.
- In the case of lecture videos, the background is not likely to be very complex, so this research uses the geometry-based method.
-
After enhancing the resolution through completion and applying binarization, the text is fed into OCR.
- The key point of the research is that increasing the resolution improves OCR performance.
-
Using the extracted text, the slides are matched.
-
To reduce computational complexity,
- Separate the slide title and body.
- OCR the title first.
- If the title is similar to the slide, then OCR the body and check.
https://dl.acm.org/doi/epdf/10.1145/957013.957080 #Reference Log