Recently I've seen this very good apps

https://play.google.com/store/apps/details?id=com.kingjim.shotnote&hl=en https://itunes.apple.com/JP/app/id411332997?mt=8

this apps reconstruct the image from the camera based on the marker (this is what I think :) ) it also have a special paper that have marker. What I've still haven't got is what is the algorithm to do this? what computer vision theory used in this one? I understand artificial intellegence and basic computer vision, is there anyone have the ide how can this be done? at first I thought this based on stereo image but based on what I've studied before stereo vision need two camera. I hope my question is clear enough but if is there any information needed please ask in the comment :)

Thank you!

  • 1
    $\begingroup$ Your link doesn't give any relevant information. What is the "marker?" on the paper? What does the app do? Does it simply make sure that the blue line on graph paper are actually equally spaced vertical and horizontal lines in the resulting jpg? You might start by looking up "Hough transform" for one way to detect the lines and then "keystone effect" for the linear transformation you need to perform to make the lines parallel. $\endgroup$ – Wandering Logic Jun 11 '13 at 12:59
  • $\begingroup$ @WanderingLogic this is an demo video -> youtube.com/watch?v=x3Oj6DGi4yU the apps scan the paper the convert it to image, maybe keystone effect can help me I'll read about it and thanks btw :) $\endgroup$ – Niko Adrianus Yuwono Jun 11 '13 at 15:47

It detects 4 markers lying on the corners of the note. One of the markers is unique to able to determine the orientation. Because the actual size and layout of the form is known, a perspective warp can be computed, bringing the scanned image to a canonical (fronto-parallel) form. Then, OCR is carried out in the region where date and time is expected. The rest of the content are taken as image and saved on the smartphone.

For OCR, character segmentation and classification might be your key words. For Marker Detection, generally black-white polygonal regions are detected and expected to be quad-formed. The rest of the blobs are filtered out and you end up with the markers.


Stereo Vision doesn't require two cameras, it only requires two views, so if there is a translation between two frames of from one camera you can apply Stereo Vision.

Having said that, the app you saw probably doesn't use any stereo information. Most apps of this kind just detect the four corners of the page (sometimes with help from the user) and warp them to a rectangle.

  • $\begingroup$ All true but this is less detailed than @WanderingLogic's comment to the question. $\endgroup$ – David Richerby Jan 6 '14 at 9:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.