We examined the performance of linear and nonlinear processors (filters) for image recognition that are l<sub>p</sub>-norm optimum in terms of tolerance to input noise and discrimination capabilities. These processors were developed by minimizing the l<sub>p</sub> norm of the filter output due to the input scene and the output due to the noise. We tested the performance of the l<sub>p</sub>-norm optimum filters by measuring the average peak-to-sidelobe ratio of the output of the filters for different values of <i>p</i>. We also tested the performance of these filters by placing a target in a scene containing additive noise and a realistic background. For the images presented here, the filters detected the target in the presence of additive noise and a realistic background. The tests conducted show that the discrimination capabilities of the l<sub>p</sub>-norm filters improve as <i>p</i> decreases (p>1). This is shown by sharper peaks at the target location and higher average peak-to-sidelobe ratios for smaller values of <i>p</i>.
© 1999 Optical Society of America
(100.5010) Image processing : Pattern recognition
Bahram Javidi and Nasser Towghi, "lp-norm optimum filters for image recognition. Part II: performance evaluation," J. Opt. Soc. Am. A 16, 2146-2150 (1999)