Showing papers by "Alan C. Bovik published in 1986"

Edge detection using median comparisons

Abstract: In this paper we describe a new edge detector based on differencing the median values of local image neighborhoods. Statistical and deterministic argument are given that suggest that this detector can be more effective than average-based detectors with similar configurations. Examples using images corrupted by Gaussian and impulse noise support the analytic results.

Optimal Detection Of Object Boundaries In Uncorrelated Speckle

Abstract: We describe several techniques for detecting object boundaries in images immersed in speckle noise. Based on the assumption that the image speckle is uncorrelated, optimal statistical procedures are formulated using local tests for changes in intensity. The first method described is parametric: the average values taken from adjacent image neighborhoods are ratioed and compared to a threshold statistic. A cooperative scheme is then described in which the parametric statistic is applied only at the zero crossings of the image resulting from a convolution with a narrowband differential operator. A non-parametric procedure based on a linear rank statistic is also described, which can be shown to be locally most powerful (among rank tests) under the noise assumption. Examples illustrate the effectiveness of each method.

Spectral analysis of order statistic filters

Abstract: We analyze the effect of filter coefficient selection on the power spectral density of an order statistic filtered signal. Assuming that the input signal is a sequence of independent and identically distributed random variates, the autocovariance and the power spectrum of the output are computed. These PSDs are compared with those of the corresponding linear finite impulse response filters with identical coefficients. It is found that, in general, low frequency components predominate regardless of coefficient selection, suggesting an inherent smoothing in the ordering process.

Detection of object boundaries in synthetic aperture radar imagery using a human visual model

Abstract: We consider the problem of automatically detecting sustained intensity changes, corresponding to object boundaries, in SAR imagery. The problem is complicated by the speckle noise found in SAR imagery, which is highly correlated and signal-dependent. In this paper, we analyze the efficacy of applying to this problem recently developed models for visual edge detection, in particular, the Marr-Hildreth Laplacian-of-a-Gaussian operator.