Jump to content
  • Advertisement
Sign in to follow this  
666_1337

Image Analysis/Filtering Question

This topic is 4526 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

Hi everybody Please excuse me if this is the wrong forum, i just thought its more graphics theory than game art... Imagine i have a sheet of recovered paper, checkered. Then I paint an image on it with a pencil, scan it up and compress it as jpeg, so i have lots of information loss about the image, and a pattern ( the checkerboard/diamonds and the artefacts of the recovered paper ) which doesn't belong to it. Is there a way to separate the pencilpainted image from the pattern, without using artificial intelligence? Thanks in advance, Alonso Schaich

Share this post


Link to post
Share on other sites
Advertisement
Are you looking for an algorithm to do it, or just general aproach?

There's some links on google dealing with similar problems ("remove textured pattern from scanned image").

Depending on how your images actually look, those methods might not be suitable.

If the pattern itself of a different color than the drawing, then you should be able to use any graphics package to filter that color out.


I imagine a FFT filter would work well for this particular case, but it really depends on how much pattern is merged with the actual drawing.

Share this post


Link to post
Share on other sites
When there seems to be a pattern with the noise of the image (all the artifacts for example) their frequencies are usually around a certain point or a couple of points when you look at the amplitude and phasis spectrum (not sure for the english terms, "spectre d'amplitude et de phase" in french)... With FFT and proper filters you could remove them and get a cleaner image.


If your using a pencil and want to extract those "patterns" you could use edge finding algorithms with the proper algorithms you can get somekind of probability for each pixel to be part of an edge. Then you chose the minimum value to reduce the edge detection (because any change in contrast may result in an edge).


I think those algorithms are presented in Numerical recipes for C/C++. Or else mayve I can try to explain them to you in a better way later (just came back from work and I'm tired).


JFF

Share this post


Link to post
Share on other sites
It shouldn't be too hard to extract the checkerboard/diamonds pattern from the image. Simply run a lowpass filter over the image to extract the pattern, then take the difference between the two images and you'll have something close to the pencildrawn image plus the wrinkles (artifacts of the recovered paper).

I'm clueless about how you'd extract the wrinkles. I don't know anything about FFT, so jff_f's solution might work. I feel like you'd have problems differentiating between the image and the wrinkles.

Share this post


Link to post
Share on other sites
Sign in to follow this  

  • Advertisement
×

Important Information

By using GameDev.net, you agree to our community Guidelines, Terms of Use, and Privacy Policy.

We are the game development community.

Whether you are an indie, hobbyist, AAA developer, or just trying to learn, GameDev.net is the place for you to learn, share, and connect with the games industry. Learn more About Us or sign up!

Sign me up!