Thursday, July 16, 2009

Image Processing in Matlab: Update 1: Filters

Okay, so I have been continuing to play around, trying to get my feet wet with image processing. I sort of jumped ahead here and begin fiddling with filters. Since I have the nadir artifact in my mosaic, I need to see if I can easily remove it.
I decided to use an adaptive filter since the noise I am most concerned is repetitive through the image (though not truly periodic as survey line spacing varies), so I do not need the same level of filtering over the whole image. I have been reading through the Matlab Image Processing documentation(pdf), and followed their directions for using "wiener2" which adaptively applies a Wiener linear filter to the image.

from the description under "help wiener2":

WIENER2 lowpass filters an intensity image that has been degraded by
constant power additive
noise. WIENER2 uses a pixel-wise adaptive Wiener
method based on statistics estimated from a local neighborhood of each

I used a 10x10 pixel neighborhood for filtering. Then, I ran the Sobel edge detect algorithm as described in my previous post. The results are below:

on the left: orginal image on the right: wiener2 filtered image. Click to see full-size image so you can see the differences.
The nadir artifacts are clearly less noticeable in the filtered image, and indeed the edge detect algorithm is now starting to pick up on the actual different patches of seafloor. The outline of the bright patch in the middle can now be seen.

I then increased the size of my filter to 20x20 pixels. Check out how well the edge detect performs now.

No comments:

Post a Comment