Week 5

Remove isolated pixels & Destriping

After we discovered that we can use the ImageVision library, we had to extend it. The way a X-Ray image goes through our application is:

Sharpen
Edge detection
Threshold
Remove isolated pixels

After that we should have a binaerised image which contains only the contours of the bones. But we learned that the original image contains stripes from scanning. If we only go through the four above steps, we get horizontal stripes in the result which makes the image unusable. That means - we have to develope a destriping class...
A class which removes the isolated pixels joined the 5 files from the third week. [Download (TGZ, 7kB)]

The source

There are only minor changes to the Makefile, view_main.cc, view_ilop.cc and view_tool.cc.
The header file contains now the class definition for remIsoPix. The out commented class definition was the first try (view_rip.cc). It works but is very slow, because it uses the get/setPixel() interface of ImageVision library. The second class definition (view_rip2.cc) descripts the recommended way. It is much faster, because it works on the image data.

The destriping problem

(added in week 10, because it is a real problem)
We are working with scaled images (less then half the size of the originals) and while doing all this image processing we discovered some nasty interfering horizontal lines. As we have to implement an contour following algorithm, which will get confused with these lines, we have to get rid of them. But how to get rid of lines without lose image information? The first few approaches I (Matthias) did, didn't work at all. (Un)Fortunally I was sick then for nearly a week, so I didn't follow them any further...

[startpage] [project part 1] [project part 2] [week 1] [week 2] [week 3] [week 4] [week 5] [week 6] [week 7] [week 8] [week 9] [week 10] [week 11] [References]

Because english is not the native language of the authors, we apologize for any inconvenience.

Last changed: Sunday, 08-Feb-2009 12:42:19 CET