Diploma work: Monthly report 2/97

Review February 1997

During the month February I was working on the basics of my prototype. I’m now able to read the single frames of my videos from January inside a GUI, apply several filters on them and generate a Delta image.

My filters always take as input and output a copy of the original image. The several filters try to detect pixels (e.g. paper, desk, hand, pen) that have to be white and turn them to white in the destination image. The source image stays untouched; so it’s the same for every instance of any filter. With every new filter-step the usability of the destination image stays at least the same as before or gets better if the filter only removes pixels that don’t belong to written text.

At the moment I apply the following filters:

1. Turn bright pixels to white. With this method most of the paper becomes a clear white without noise.
2. Analyse areas (about 9x9 to 13x13 pixels) around a pixel: If the ratio of dark pixels around such a pixel is above a certain limit the pixel is part of a shadow, the hand or the desk and is eliminated to white. If the ratio is below a certain level I assume the pixel to be part of the written text an leave it untouched. With this method most of the shadows and the hand is eliminated to a clear white.
3. Black and White conversion

The results of the work during February are the following ones:

• Routines to display my digital images (True Color) under X11. With those routines I’m able to display True-Color images either on a True-Color display (e.g. LINUX PC at home) and on a 256 color display (e.g. SUN Workstation at University)
• The filters allow me to generate a rather good base-image to produce the delta frames. The heaviest problem at the moment are the small shadows around the pen tip. Those shadows are rather dark and small. Because those shadows are only about 3 pixels thick and the written text has about the same thickness I detect those shadows as text strokes and don’t turn them to white as they should be.
• Generation of delta images out of greyscale images.

Preview March 1997

In the next month I still focus my work on the processing of digital images and on the generation of the delta images. At the end of March I plan to finish my prototype. Concerning my Roadmap this prototype has to be able to process images inside a GUI and deliver some rather good output in my own data format.

To achieve this I have to concentrate on the following points:

• Elimination of the shadows around the pen tip during the capturing with the camera. I plan to tune the illumination to reduce those shadows or at least turn them to a brighter greyscale.
• Elimination of the shadows around the pen tip using a filter. I’ll try to tune the existing ratio-based filter or have to implement a new one.
• Specification of my own data format to store the recognized linestrokes
• Design and implementation of a useful GUI (X11 and XView/Motif)
• Tuning the speed of the filters. Mainly the filters that treat a region around a pixel take a long time to process.

zurück zur Uebersicht der Diplomarbeit

Last modified 23.12.1999 00:18:50 by Thomas von Siebenthal