Using interpolation to restore corrupted image data?

2 views (last 30 days)
So I'm trying to restore corrupted image data using interpolation and am having less than zero luck. I have tried a bunch of different things and none of them seem to help. Here's what I have right now which generates an image but it looks exactly the same as the original. Any suggestions? (The image data was given in a 195x1099 double)
paint=SanRomano;
sizeP=size(paint);
x=1:sizeP(1);
y=1:sizeP(2);
figure(3);
[xGrid,yGrid]=meshgrid(1:1:1099,1:1:195);
[xS,yS]=meshgrid(1:0.5:1099,1:0.5:195);
vq=interp2(xGrid,yGrid,paint,xS,yS,'cubic');
image(vq)

Answers (2)

Jan
Jan on 23 Mar 2017
An interpolation is only useful for the restauration of an image, if the damaged pixles are identified at first an then replaced by the interpolation of the neigboring pixels. Currently your interpolation inserts only pixels between the existing pixels with their interpolated values. This means that you get the corrupted image with a finer resolution, but no substantial changes.
  2 Comments
Ella Kaplan
Ella Kaplan on 23 Mar 2017
Is there a different method that would result in substantial changes?
Stephen23
Stephen23 on 23 Mar 2017
@Ella Kaplan: you need to identify the "corrupted" pixels, and interpolate new values at those positions. The interpolation is the easy part. Identifying the corrupt pixels is the hard part: start by using google scholar.

Sign in to comment.


Walter Roberson
Walter Roberson on 23 Mar 2017
Edited: Walter Roberson on 23 Mar 2017
If you replace the damaged pixels with nan then you could use John D'Errico file exchange contribution inpaint_nan

Categories

Find more on Convert Image Type in Help Center and File Exchange

Products

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!