Two-dimensional adaptive enhancement and detection of linear features in noisy images
Curtis Knittle, Neeraj Magotra · 2002
The authors present results from using a novel two-dimensional adaptive filtering algorithm to enhance linear features embedded in noisy images. The algorithm is called the two-dimensional adaptive correlation enhancer (2DACE). The 2DACE algorithm is a coefficient update scheme that converges to the two-dimensional autocorrelation function of the signal or feature of interest. Since the noise statistics are nonstationary and the duration of a signal or feature can be very short, the algorithm must be capable of rapid convergence to its optimal impulse response, a requirement that 2DACE meets. After the image has been enhanced, a modified Radon transform is employed to explore the improved detectability of the linear features. It is shown that preprocessing the image with 2DACE prior to taking the transform yields a great improvement in signal-to-noise ratio in the transform domain. It is also shown that the adaptive filter significantly enhances surface ship wakes in SAR (synthetic aperture radar) images.>