Last Update 29 January 2002

Welcome to the Optimark homepage. Optimark is a benchmarking tool for still image watermarking algorithms that was developed in the Artificial Intelligence and Information Analysis Laboratory at the Department of Informatics, Aristotle University of Thessaloniki, Greece. Its main features can be summarized as follows:

• Graphical user interface.
• Detection/decoding performance evaluation using multiple trials utilizing different watermaking keys and messages. 
• Evaluation of the following detection performance metrics:
  • For watermark detectors that provide a float output, i.e., the value of the test statistic used for detection:
    • Receiver Operating Characteristic curves (ROC), i.e. plots of the probability of false alarm versus the probability of false rejection.
    • Equal Error Rate.
    • Probability of false alarm for a fixed, user defined, probability of false rejection.
    • Probability of false rejection for a fixed, user defined, probability of false alarm.
  • For watermark detectors that provide a binary output, i.e. a value that states whether the watermark has been detected or not:
    • Probabilities of false alarm and false rejection.
• Evaluation of the following decoding performance metrics, for algorithms that allow for message encoding (multiple bit algorithms):
  • Bit error rate.
  • Percentage (probability) of perfectly decoded messages.
• Evaluation of the mean embedding and detection time.
• Evaluation of the algorithm payload (for multiple bit algorithms).
• Evaluation of the algorithm breakdown limit for a certain attack and a certain performance criterion, i.e., evaluation of the attack severity where algorithm performance exceeds (or falls below) a certain limit.
• Result summarization in multiple levels using a set of user defined weights on the selected attacks and images.
• Option for both user defined and preset benchmarking sessions.

The attacks that are currently included in Optimark are the following:

1. No Attack
2. Cropping
3. Line and Column Removal
4. General Linear Transformation
5. Scaling
6. Shearing
7. Horizontal Flip
8. Rotation
9. Rotation + Autocropping
10. Rotation + Autocropping + Autoscale
11. Sharpening
12. Gaussian Filtering
13. Median
14. Jpeg

The above attacks have been re-programmed from the Stirmark benchmark.

If you use the Optimark tool for your research, please cite the following paper: 
V. Solachidis, A. Tefas, N. Nikolaidis, S. Tsekeridou, A. Nikolaidis, I.Pitas, ``A benchmarking protocol for watermarking methods’’, 2001 IEEE Int. Conf. on Image Processing (ICIP'01), pp. 1023-1026, Thessaloniki, Greece, 7-10 October, 2001 
Reading this paper before using Optimark is highly recommended.

Please send your comments and bug reports to optimark@zeus.csd.auth.gr

Optimark was partially supported by EU Projects CERTIMARK & INSPECT.