Diploma work: Monthly report 4/97
Review April 1997
During April I spent my time with the analysis of the delta
images, the preparation of my postponed talk and the problems of
abstract base classes in C++. Because I had to do a lot of other
things besides my diploma work I didn’t get as far as
expected. The the results of April are:
- I implemented an abstract base class Image with two
son’s, called Image8 (Greyscale Images) and Image24
(True Color Images). All my classes and filters work now
seamlessly on either greyscale or true color images. The
speed of the filters is, after some tuning, now
acceptable. In the current relase of my programm all the
filters work on greyscale images.
- To get the position of the pen in a image I calculate the
average position of all the black pixels in a delta image
and write this position to my file. If there are no black
pixels in a delta image the pen was up and an empty
position is written to the output file.
- First Preparation of my talk (which was postponed to
June)
| The output generated by my prototype looks like this:
|
Output: |
 |
| |
Input: |
 |
Because the generation of the delta images is not yet good
enough, about half of the line strokes are lost.
Preview May 1997
The generation of the delta images works more or less. Because
sometimes the brightness of an ink trace is very near to the
brightness of the paper or a shadow I have to choose the limit
between black and white very conservative and loose some ink
traces. But I don’t get to many ink trace. On the other hand
I could allow more ink traces to be detected, but that would
result in too many ink positions ...
My goals for May 1997 are:
- Improve the generation of the delta images to
differenciate better between growing ink traces and
vanishing or moving non-ink pixels.
- Improve the preprocessing step to achieve a better result
with the delta image generation. The existing filters to
reduce the hand and bright paper background don’t
help much, because there are a lot of points on the
border between the hand and the paper which are not
eliminated out of the image. And exacltly those points
are sometimes detected as wrong ink pixels.
- With the two improved algorithms from above I should get
good results on the existing image sequences.
- Read about projective geometry to be able to reproject
the captured image into a normalized view of the paper.
With this I should be able to detect the position of the
paper and normalize into the point of view of a scanner
or a graphics tablet.
zurück zur Uebersicht der
Diplomarbeit
Last modified 23.12.1999 00:18:51 by Thomas von Siebenthal