Segmentation Demo

Intelligent Segmentation

The following page contains an example of the intelligent segmentation process in action.

A downloadable DEMO (Segdemo.zip) of our earlier algorithm is also available. The Demo program (segdemo.exe) is a 16-bit DOS application that should run under any WINDOWS-based platform.

Step 1. The Original Image

The figure below shows a handwritten word obtained from the CEDAR CD-ROM.

Figure 1. Original Image

Step 2. Heuristic Segmentation

The original image is now over segmented using a heuristic segmentation algorithm [1].

Figure 2. Segmented image using heuristic algorithm

Step 3. Intelligent Segmentation

Finally, the segmentation points generated by the heuristic algorithm are presented to a trained ANN. The ANN classifies these points into "correct" and "incorrect". Finally all "correct" points are left untouched and all "incorrect points are removed.

Figure 3. Image after incorrect points are removed

Following the segmentation process, individual characters can be recognised using an ANN. Finally, words can be recognised using a neural-based lexicon [2].

References

[1] M. Blumenstein, and B. Verma, "A Neural Based Segmentation and Recognition Technique for Handwritten Words", World Congress on Computational Intelligence (WCCI '98), Anchorage, Alaska, 1738-1742.

[2] B. Verma, M. Blumenstein, and S. Kulkarni, "Recent Achievements in Off-line Handwriting Recognition Systems", International Conference on Computational Intelligence and Multimedia Applications (ICCIMA '98), Melbourne, Australia,27-33