Finding Braille(embossed printing) in an image
in [Matlab]
|
Prev: Finding Braille(embossed printing) in an image
Next: Warning: Texture data too large... problem with alphadata in imagesc
From: David Elsdoerfer on 11 Aug 2010 11:31 Hi folks! I am currently working on a project for my image processing classes in my university. The task is to find braille located on a drugbox. I already managed to separate the braille from the rest of the image but now I'm stuck searching for a way to separate the 2 by 3 dots patterns, reading their meanings and returning the proper text in alphabetical characters. For my process so far see http://drop.io/dkh07pe . Thanks in advance for your help.
From: us on 11 Aug 2010 11:38 "David Elsdoerfer" <david.elsdoerfer(a)gmx.net> wrote in message <i3ufo7$egd$1(a)fred.mathworks.com>... > Hi folks! > I am currently working on a project for my image processing classes in my university. The task is to find braille located on a drugbox. I already managed to separate the braille from the rest of the image but now I'm stuck searching for a way to separate the 2 by 3 dots patterns, reading their meanings and returning the proper text in alphabetical characters. For my process so far see http://drop.io/dkh07pe . Thanks in advance for your help. do NOT double post(!)... CSSMers don't respond well to this kind of behavior... us
From: us on 11 Aug 2010 11:45 "David Elsdoerfer" <david.elsdoerfer(a)gmx.net> wrote in message <i3ufmd$ap2$1(a)fred.mathworks.com>... > Hi folks! > I am currently working on a project for my image processing classes in my university. The task is to find braille located on a drugbox. I already managed to separate the braille from the rest of the image but now I'm stuck searching for a way to separate the 2 by 3 dots patterns, reading their meanings and returning the proper text in alphabetical characters. For my process so far see http://drop.io/dkh07pe . Thanks in advance for your help. very nice image, indeed... now - where is your ML language related question(?)... us
From: David Elsdoerfer on 11 Aug 2010 11:51 "us " <us(a)neurol.unizh.ch> wrote in message <i3ug4s$alb$1(a)fred.mathworks.com>... > "David Elsdoerfer" <david.elsdoerfer(a)gmx.net> wrote in message <i3ufo7$egd$1(a)fred.mathworks.com>... > > Hi folks! > > I am currently working on a project for my image processing classes in my university. The task is to find braille located on a drugbox. I already managed to separate the braille from the rest of the image but now I'm stuck searching for a way to separate the 2 by 3 dots patterns, reading their meanings and returning the proper text in alphabetical characters. For my process so far see http://drop.io/dkh07pe . Thanks in advance for your help. > > do NOT double post(!)... > CSSMers don't respond well to this kind of behavior... > > us Sorry I don't understand... What do you mean by double post? I typed my question and then hit the "Post Mesage" button. I've never posted in a forum before so if I made some beginners mistake I hereby apologize.
From: Sean on 11 Aug 2010 11:54
"David Elsdoerfer" <david.elsdoerfer(a)gmx.net> wrote in message <i3ufmd$ap2$1(a)fred.mathworks.com>... > Hi folks! > I am currently working on a project for my image processing classes in my university. The task is to find braille located on a drugbox. I already managed to separate the braille from the rest of the image but now I'm stuck searching for a way to separate the 2 by 3 dots patterns, reading their meanings and returning the proper text in alphabetical characters. For my process so far see http://drop.io/dkh07pe . Thanks in advance for your help. I don't really know anything about braille but here are a few ideas: -Are all letters (or words?) the same size? I.e. can you make a grid that will segment the each block of dots into it's own block. -If you can do this first and run the rest of the algorithm on each block. I would work really hard to be able to do this. -Count dots. So say 10 letters have 5 dots and your image has 5 objects (bwconncomp()); you've already limited it to these options. -Use a minimum absolute difference as your measurement algorithm. This will be difficult to implement, but then again this isn't a trivial task! -Have each pattern saved in a matrix. You'll have to up or down sample this matrix to be the same size as the blocks above. The blocks above should be squeezed in to their matrix so that each border has a true value on it. -Calculate absolute difference. I.e. subtract the pattern from you predefined file from your image and take the absolute value of the sum of all voxels. -Whichever pattern has the minimum should be the winner. Return that letter. Good Luck! -Sean |