Diploma work: Monthly report 7/97

Review July 1997

I did the following things during the last month:

For the detection of the pen position out of a delta image I implemented two new averaging algorithms. The first algorithm calculates the center with a weight proportional to the darkness of a pixel. This algorithm can be also applied iterative. The other algorithm calculates the center as the median of the x and y coordinates. The test showed that the median worked good and the greyscale worked well. Both worked rather robust all over the image sequencies.
To improve the resulting pen positions I used interpolation. If there’s a gap between two "pen downs" of 1 or 2 "pen ups" and the distance is not too long I linear interpolate the intermediate pixels. Besides two neighbouring pixels with a too long distance aren’t connected with a line.
I changed some small things at the GUI. It’s now much easier to test image sequencies and I get more informations during the processing phase. The processing can now be automatic, too.
I tried to get more informations out of the Otsu parameters, but I couldn’t find coherences between the parameters (probabilities, seperability, ...) and the apperance of the Histogram (Hills, ..).
I tried to detect pen positiona out of a new image sequence with a general pen. I used a pen with a black pen tip and a black front part to write a line of text. The results weren’t useful.
Most problems at the moment are caused by letters which have parts that are covered during writing and freed again. Problematic letters are e.g. "e", "o", "h".

Two examples:	Detected Linestrokes:
	Pixels used for the detection:

My goals for August 1997 are:

At the moment I have a memory leak in my program which consumes about 1MB of RAM for every image. This prevents me from detecting larger sequencies than 250 images. But it shouldn’t be a big problem to solve this using Purify.
I will capture some more sequencies with pens that have different ink colors (red, green, grey) and with other writers than myself.
I have to improve the calculation of the center with the detected delta pixels. The both approaches with greyscales and median are good, but a combination could be even better.
I will improve the combination of the detected ink positions to line strokes. Here I see a lot of possibilities to improve the results.
Analyse the influence of the ETHZ input interface to my detection process and my output interface.
Think about the automatic getting of the mask after the writer completed a line of text.
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.

Last modified 23.12.1999 00:18:51 by Thomas von Siebenthal