Formal system specification and testing of image processing/computer vision algorithms

Andrew C. Segal; John Quintas; Robert Kero; Richard Greene; Greg Chisholm

doi:10.1117/12.149064

Translator Disclaimer

4 August 1993 Formal system specification and testing of image processing/computer vision algorithms

Andrew C. Segal; John Quintas; Robert Kero; Richard Greene; Greg Chisholm

Author Affiliations +

Proceedings Volume 1909, Device-Independent Color Imaging and Imaging Systems Integration; (1993) https://doi.org/10.1117/12.149064
Event: IS&T/SPIE's Symposium on Electronic Imaging: Science and Technology, 1993, San Jose, CA, United States

Abstract

An important issue in the advancement of Image Processing and Computer Vision (IP/CV) algorithm development is the ability to test, verify and compare the newly developed algorithm performance to other functionally equivalent algorithms and to ground truth in a complete and meaningful way. In this paper, we explore several criteria to estimate the performance of IP/CV algorithms against databases of synthetic and real data and use statistical analysis to determine performance. Formal system analysis along with Monte Carlo testing is used to rigorously compare these algorithms. A case study is performed on the familiar edge detection algorithms to illustrate the techniques and also the vulnerability of these algorithms to good average performance but in certain cases to very poor performance.

Citation Download Citation

Andrew C. Segal, John Quintas, Robert Kero, Richard Greene, and Greg Chisholm "Formal system specification and testing of image processing/computer vision algorithms", Proc. SPIE 1909, Device-Independent Color Imaging and Imaging Systems Integration, (4 August 1993); https://doi.org/10.1117/12.149064

ACCESS THE FULL ARTICLE