hi, everyone i am working on script identification from multi-lingual document image, i need to know which is the best feature extraction method to extract features from grayscale document image.
thanking you,
With regards,
SATISH KUMAR
Hi, I will not disapoint you but there is no 'best method'. Many aproaches are available. However, it depends on the concrete application. You have to ask youself what kibd of features are necessary for your tasks (sematically). Then it is possible to discuss different approaches.
Thanks for the suggestion Volker Lohweg, in my view the best method is that extracting the prominent features from an image, which is capable of making the discriminating with other scripts. i have used LBP, GLCM, now i need other feature extraction method. for example i have given the images from which i need to extract features.
Thanks.
From, SATISH KUMAR
You need to identify what are the key differences between the scripts.
As for an example the second script seems to have lots of horizontal lines, compared to the other two. so you can use hough transform (may be to detect lines) to discriminate it.
As an overall idea I would say
SIFT,SURF, Histogram of oriented gradients are good feature extraction mechanisms.
Take a look here. Usually Google Research is a good source of information on your topic.
https://static.googleusercontent.com/media/research.google.com/de//pubs/archive/35506.pdf
Good luck,
Volker
Thank you, Chamri Silva and Volker Lohweg. i consider the both suggestions.
Is LBP and HOG features are better for above shown document image blocks?
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Human-Computer Interaction