size normalization or scaling for a drawn graph in x,y coords format

Hi everyone , I'm a beginner in matlab ..my project is Online Arabic handwriting recognition , I have drawn the character on an axes then a text file - which contains the x and y coords of it - is saved , then I need to apply some preprocessing steps like translation , scaling , resampling and I want to know how can I make scaling .. I tried to use interp and interp1 functions but when I plot their result It doesn't seem right.

if the interpolation is used in scaling then how can I use it? what functions I've to use ? if not , how can I do the scaling?

Thanks

Answer 1

[Image Analyst]

Are you following one of the published methods here: http://iris.usc.edu/Vision-Notes/bibliography/contentschar.html#OCR,%20Document%20Analysis%20and%20Character%20Recognition%20Systems Look at Section 23.4.6.5 Arabic Numbers, Digits, Handwritten, Numberal Recognition. Those papers will probably tell you if you need to do that or not.

For scaling you can use imresize.

[aurora]

Thank you for your response, imresize is for resizing an image and that's useful for offline recognition but not for online recognition, in online recognition the data(character) isn't an image but a sequence of x & y coordinates.

my project isn't related to image processing

[Image Analyst]

Perhaps we should alert the OCR community. They've been processing the x,y coordinates of the "written" material for years, and evidently they've been wasting their time all along.

[aurora]

excuse me! I didn't understand what you mean .. I also need to know how to apply re-sampling , interpolation and smoothing for the x,y coordinates that are saved in a text file as two columns when I implement the matlab functions of (resample) or (interp) or (smooth) the result plotted isn't as I want .. the character shape changes .

[Image Analyst]

I have little to no experience in OCR algorithms so I don't know if what you propose to do would be the best approach or not. That's why I referred you to the experts who know - those who have published successful algorithms on the topic. I think you will find your answer there. Good luck.