|
Computer vision is my special field of interest. Here are some of my experiments in this field.
3D-reconstruction using a modified silhouette method: This was my first project on computer vision. It is a few years old and I wouldn't do it this way again. But at that time,
I thought it might be a good idea. Have a look at this page. I wrote it in german those days. If you can't read it, try altavista translator or just have a look at the colorful pictures. :)
|
|
Here are some screenshots of my first feature tracker. Neat, isn't it?
|
|
This was my first structure-from-motion approach. The program captures pictures from a webcam (1st window on the topleft), removes lensdistortion (2nd window), tracks some features (3rd window).
The internal parameters were calibrate before so undistortion of the pictures can be done right after capturing. After capturing some pictures, the program begins to calculate the camerapositions an 3D-positions of the tracked features (note the dark blue points in front of the cyan lines).
I used the 8-point algorithm combined with a least square and randac method to compute the initial guess and afterwards optimize it nonlinear.
There was a problem with my webcam those days producing only red pictures. Note the well selected reference object(s) I captured. :)
|
|
Lens & shadingcorrection This was a job offered by "Lehrstuhl für Regelungstechnik und Signaltheorie" from the faculty of eletrical engenering.
They searched for someone to write a program for Pentacon that removes lens and shading-errors of captured images by a referenceimage.
The first image is the reference captured by the camera. The second is the picture that has to be corrected. The third is the result. The pictures are scaled down, original size was about 10000x7000.
|
|
|