What should be the standard value of noise in Jtessbox editor and what does the field letter tracking indicate? What is its standard value? I should make an OCR than can read bad images as well.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I actually read the documentation but the problem is that the images generated by Jtess box editor are too high quality thus the traindata cannot work on low quality images, but if I make a traindata with low quality images, it can work on both low and high quality images, in that case, adding noise may help but I did not find any difference in the image.So, I wanted to know the value.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
What should be the standard value of noise in Jtessbox editor and what does the field letter tracking indicate? What is its standard value? I should make an OCR than can read bad images as well.
Letter tracking is space between letters (letter-spacing).
Some developers indicated that adding some noise would increase recognition rates.
https://groups.google.com/forum/#!searchin/tesseract-ocr/noise/tesseract-ocr/mlcmD_dSClo/7RFyku9MDvcJ
There are no standard values. You'd have to fiddle with them to find the optimal values for your training.
PS. Apparently, you did not read the documentation. All of this information has been documented on the website.
Last edit: Quan Nguyen 2015-05-11
I actually read the documentation but the problem is that the images generated by Jtess box editor are too high quality thus the traindata cannot work on low quality images, but if I make a traindata with low quality images, it can work on both low and high quality images, in that case, adding noise may help but I did not find any difference in the image.So, I wanted to know the value.