Showing papers by "Richard C. Wilson published in 2000"

A Bayesian framework for radar shape-from-shading

Abstract: This paper introduces a Bayesian approach to shape-from-shading (SFS) which is applied to terrain recovery in synthetic aperture radar (SAR) images. The Bayesian model relates the recovery of 3-D shape information to the original 3-D radar intensity and to edges separating different topographic regions. First, we model the image amplitude distribution and the reflection function in SAR images. Using a maximum log-likelihood feature detector derived from the image statistics we identify the ridges and ravines in the terrain image. These topographic features are used to constrain the recovery of surface normals in the shape-from-shading process. Finally, the surface normals are smoothed using robust statistics operators.

Tangent Fields from Population Coding

Abstract: This paper addresses the problem of local orientation selection or tangent field estimation using population coding. We use Gabor filters to model the response of orientation sensitive units in a cortical hypercolumn. Adopting the biological concept of population vector decoding [4], we extract a continuous orientation estimate from the discrete set of responses in the Gabor filter bank which is achieved by performing vectorial combination of the broadly orientation-tuned filter outputs. This yields a population vector the direction of which gives a precise and robust estimate of the local contour orientation. We investigate the accuracy and noise robustness of orientation measurement and contour detection and show how the certainty of the estimated orientation is related to the shape of the response profile of the filter bank. Comparison with some alternative methods of orientation estimation reveals that the tangent fields resulting from our population coding technique provide a more perceptually meaningful representation of contour direction and shading flow.

Terrain feature identification by modeling radar image statistics

Abstract: We propose a new statistical model for SAR images. According to this model, the SAR image amplitude follows a product of Rayleigh and Bessel functions. We derive the maximum likelihood feature detector for extracting terrain features from synthetic aperture radar (SAR) images. The terrain features are classified as ridges and ravines according to their statistical properties and surrounding neighborhood. These salient features are used as constraints for estimating the SAR terrain surface.

3-D terrain from synthetic aperture radar images

Abstract: Surface analysis is important for automatic terrain cartography and for airborne navigation. This paper proposes a new approach to shape-from-shading (SFS) in synthetic aperture radar (SAR) images. The SFS problem is embedded in a Bayesian framework. We maximize the surface orientation probability using SAR image statistics, local smoothing and constraints imposed by object discontinuities. We model the statistics of the SAR image distribution as a product between the Rayleigh and Bessel functions. We derive the optimal edge detector for this distribution. The resulting edges are classified as ridges and ravines according to a statistical test. Afterwards, the edges are used as constraints in the estimation of the surface normals. We propose various smoothing algorithms for the vector field of surface normals using robust statistics and surface curvature consistency. The results provided by these algorithms are compared with those given by local averaging.

Population codes for orientation estimation

Abstract: Population coding has become an essential paradigm in cognitive neuroscience over the past decade and is increasingly studied within the neural network community. We investigate the use of population vector decoding for local edge orientation estimation from a discrete set of Gabor filters. Vectorial combination of the broadly tuned filter outputs yields a resultant population vector, which gives a precise and robust estimate of the local contour orientation. We present results on the accuracy and robustness of orientation measurement.

Terrain modeling in synthetic aperture radar images using shape-from-shading

Abstract: We introduce a new approach for recovering shape-from-shading (SFS) from synthetic aperture radar (SAR) images of the terrain. Three contributions are proposed: 1) we show how the direction of surface normals is constrained by the geometry of the radar reflectivity cone; 2) we show how topographic features can be used as boundary constraints on the recovered surface normals; and 3) the resulting field of surface normals is smoothed using robust statistics.

Storage capacity of the exponential correlation associative memory

Abstract: We analyze the pattern storage capacity of the exponential correlation associative memory (ECAM). We model the performance of the ECAM when presented with corrupted input patterns. Our model leads to an expression for the storage capacity of the ECAM both in terms of the length of the bit-patterns and the probability of bit-corruption in the original input patterns. These storage capacities agree closely with simulation. In addition, our results show that slightly superior performance can be obtained by selecting an optimal value of the exponential constant.

Decoding population codes

Abstract: Population coding is a coding scheme used in a neural systems and is of general importance. It is ubiquitous in neurological systems. For this reason there is great interest in exploiting population coding in pattern recognition algorithms. A population of neural activities represents not only the value of some variable in the environment, but a full probability distribution for that variable. The information is held in a distributed and encoded form which may in some situations be more robust to noise and failures than conventional representations. Encoding a population code with discrete-valued elements creates inaccuracies in the coded distributions. The result of these errors is the introduction of spurious high-frequency noise in the final distribution. We develop two methods of eliminating these errors and present results comparing the reconstruction accuracy of these techniques.

Optimising pattern recovery in recurrent correlation associative memories

Abstract: Addresses the problem of how to identify the optimal excitation function for the recurrent correlation associative memory. We present a model of pattern recovery which allows us to measure probability of bit-error. By minimising this measure we are able to numerically locate the excitation function which results in the minimum error of pattern recall. Additionally, we show that minimising a simpler measure of pattern overlap leads to an analytical expression for the excitation function which is exponential. We compare the performance of the numerical and exponential functions. This reveals that the more easily controlled exponential is only slightly poorer in its performance.

Terrain feature classification in SAR imagery

Abstract: This paper describes a maximum likelihood feature detector for extracting terrain features from Synthetic Aperture Radar (SAR) images. We commence by deriving the probability distribution for SAR image amplitude. According to our model, the SAR image amplitude follows a product of Rayleigh and Bessel functions distribution. We show how the two parameters of this model can be estimated using robust statistics. With the model to hand we develop a maximum likelihood feature detector. Eventually, we classify the detected terrain features in ridges or ravines.