From FNS to HEIV: a link between two vision parameter estimation methods

W. Chojnacki, M. J. Brooks, A. van den Hengel, D. Gawley
Submitted.
PDF (Link-02.pdf, 0.11Mb)

Problems requiring accurate determination of parameters from image-based quantities arise often in computer vision. Two recent, independently developed frameworks for estimating such parameters are the FNS scheme of the authors, and the HEIV scheme of Leedan \& Meer. In this paper, it is shown that the two schemes constitute intimately related but different means of numerically solving a common underlying equation characterising the minimiser. The analysis is driven by the search for a non-degenerate form of a certain generalised eigenvalue problem, and this effectively leads to a new derivation of the HEIV algorithm. This work may be seen as an extension of the authors' previous efforts to rationalise and inter-relate a spectrum of estimators, including the renormalisation method of Kanatani and the normalised eight-point method of Hartley.
 


Revisiting Hartley's normalised eight-point algorithm

W. Chojnacki, M. J. Brooks, A. van den Hengel, D. Gawley
Submitted.
PDF (normalisation-02.pdf, 0.14Mb)

The paper gives a novel explanation for the improvement in performance of the stereo-vision eight-point algorithm that results from using normalised data. It is first established that the normalised algorithm acts to minimise a specific cost function. It is then shown that this cost function is statistically better founded than the cost function associated with the non-normalised algorithm. This supercedes the standard agument that improved performance is due to the better conditioning of a pivotal matrix. Experimental results are given that support the shift in argument from numerical stability to statistical soundness as a means of rationalising performance. This work continues a wider effort to place a variety of estimation techniques within a coherent framework.
 


A new constrained parameter estimator: experiments in fundamental matrix computation

A. van den Hengel, W. Chojnacki, M. J. Brooks, D. Gawley
British Machine Vision Conference 2002.
PDF (BMVC2002.pdf, 0.19Mb)

In recent work the authors proposed a wide-ranging method for estimating parameters that constrain image feature locations and satisfy a constraint not involving image data. The present work illustrates the use of the method with experiments concerning estimation of the fundamental matrix. Results are given for both synthetic and real images. It is demonstrated that the method gives results commensurate with, or superior to, previous approaches, with the advantage of being fast.
 


A new approach to constrained parameter estimation applicable to some computer vision problems

W. Chojnacki, M. J. Brooks, D. Gawley, A. van den Hengel
In D. Suter, editor, Statistical Methods in Video Processing Workshop held in conjunction with ECCV'02, Copenhagen, Denmark, June 1-2, 2002.
PDF (SMVP2002.pdf, 0.28Mb)

Previous work of the authors developed a theoretically well-founded scheme (FNS) for finding the minimiser of a class of cost functions. Various problems in video analysis, stereo vision, ellipse-fitting, etc, may be expressed in terms of finding such a minimiser. However, in common with many other approaches, it is necessary to correct the minimiser as a post-process if an ancillary constraint is also to be satisfied. In this paper we develop the first integrated scheme (CFNS) for simultaneously minimising the cost function and satisfying the constraint. Preliminary experiments in the domain of fundamental-matrix estimation show that CFNS generates rank-2 estimates with smaller cost function values than rank-2 corrected FNS estimates. Furthermore, when compared with the Hartley-Zisserman Gold Standard method, CFNS is seen to generate results of comparable quality in a fraction of the time.
 


What value covariance information in estimating vision parameters?

M. J. Brooks, W. Chojnacki, D. Gawley, A. van den Hengel
International Conference on Computer Vision, Vancouver, July 2001.
PDF (iccv01.pdf, 0.118Mb)

Many parameter  estimation methods used in computer  vision are able to utilise covariance information describing the uncertainty of data measurements.  This paper considers the value of this information to  the  estimation  process  when   applied  to  measured  image  point  locations. Covariance  matrices are first described  and a procedure  is then  outlined whereby covariances  may be associated  with image  features located  via a measurement process.  An  empirical study is  made of  the conditions  under which covariance  information enables  generation of  improved parameter  estimates.  Also explored  is the  extent to which the noise should be anisotropic and inhomogeneous if  improvements  are  to  be  obtained over  covariance-free  methods.  Critical  in this  is the  devising of  synthetic  experiments under  which  noise conditions  can  be precisely  controlled.  Given  that  covariance information  is, in itself, subject  to estimation error,  tests  are also  undertaken  to determine  the  impact of  imprecise  covariance  information upon  the  quality of  parameter estimates.  Finally,  an  experiment is  carried  out  to  assess the  value  of  covariances in estimating the fundamental matrix from real images.
 


Rationalising the Renormalisation Method of Kanatani

W. Chojnacki, M. J. Brooks, A. van den Hengel
Journal Mathematical Imaging and Vision, 14, 1, 2001, 21-38.
PDF (jmiv.pdf, 110Kb)

The renormalisation technique of Kanatani is intended to iteratively  minimise a cost function of a certain form while avoiding systematic  bias inherent in the  common method of minimisation  due to Sampson.   Within the   vision  community,  the  technique has   generally been  perceived as  somewhat  controversial and impenetrable.   This  work  presents an alternative, simpler derivation  of the technique, along  with new insights that place it in the context  of other approaches.   We first show that the minimiser of the cost function must satisfy a  special variational equation.   A Newton-like, fundamental numerical  scheme is presented with  the  property that its theoretical   limit  coincides with  the minimiser.  Standard statistical  techniques are  then employed to derive afresh several renormalisation schemes.  The  fundamental  scheme  proves pivotal  in   the rationalising of  the renormalisation  and other schemes, and  enables us to show that the  renormalisation schemes do not  have as their theoretical limit the  desired minimiser.  The various minimisation schemes are finally  subjected to a rigorous performance analysis.


On the fitting of surfaces to data with covariances

W. Chojnacki, M. J. Brooks, A. van den Hengel, D. Gawley
IEEE Trans. Pattern Analysis Machine Intelligence, 22, 11, Nov. 2000, 1294-1303.
PDF (pami.pdf, 0.22Mb)

We consider the problem of estimating parameters of a model described by an equation of special form.  Specific models arise in the analysis of a wide class of computer vision problems, including  conic fitting and estimation of the fundamental matrix.  We assume that noisy data are accompanied by (known) covariance matrices characterising the uncertainty of the measurements.  A cost function is first obtained by considering a maximum likelihood formulation, and applying certain necessary approximations that render the problem tractable. A novel, Newton-like iterative scheme is then generated for determining a minimiser of the cost function.  Unlike alternative approaches such as Sampson's method or the renormalisation technique, the new scheme has as its theoretical limit the minimiser of the cost function.  Furthermore the scheme is simply expressed, efficient, and unsurpassed as a general technique in our testing.  An important feature of the method is that it can serve as a basis for conducting theoretical comparison of various estimation approaches.


Fundamental matrix from optical flow: optimal computation and reliability evaluation

K. Kanatani, Y. Shimizu, N. Ohta, M.J. Brooks, W. Chojnacki, A. van den Hengel
Journal of Electronic Imaging, 9, 2, April 2000, 194-202.
PDF (jei.pdf, 498Kb)

The optical flow observed by a moving camera satisfies, in the absence of   noise, a special equation  analogous   to the epipolar constraint arising in  stereo vision.  Computing  the ``flow fundamental matrix'' of   this  equation  is   an   essential prerequisite  to  undertaking 3-D analysis  of the   flow.   This paper presents   an optimal formulation of the problem of estimating  this matrix under an assumed noise model.  This model admits independent Gaussian noise that is not necessarily isotropic or homogeneous.  A  theoretical bound is derived for the  accuracy of the estimate.   An algorithm is then devised that employs a  technique called renormalization to deliver an estimate and then corrects the estimate so as  to satisfy a particular decomposability condition.   The algorithm also provides an evaluation of the reliability of  the  estimate.  Epipoles and  their  associated reliabilities are  computed   in    both  simulated   and   real-image experiments.  Experiments indicate that the algorithm delivers results in the vicinity of the theoretical accuracy bound.


Fitting surfaces to data with covariance information: 
fundamental methods applicable to computer vision

W. Chojnacki, M. J. Brooks, A. van den Hengel
TR99-03, Department of Computer Science, University of Adelaide, August 1999.
PDF (TR03.pdf, 137Kb)

We are concerned with solving an equation whose form is applicable to a wide class of problems arising in computer vision. The equation typically relates image point locations to the parameters of some appropriate model. We assume that each measured datum is accompanied by a covariance matrix that characterises the uncertainty of the measurement. Noisy data are assumed to be in plentiful supply, implying that our problem is overdetermined. To tackle noise, the problem is transformed to one of least squares minimisation. In this sense, we are concerned with fitting a surface to data and their covariances. Examples are given of computer vision problems whose forms constitute instances of our general equation. The paper has two principal concerns: the establishing of a suitable cost function for our general problem, and the deriving of effective schemes for minimising the cost function. A weighted least squares (WLS) cost function is obtained by considering an optimal maximum likelihood formulation, and applying certain necessary approximations that render the problem tractable. A new and fundamental Newton-like iterative scheme is then generated for directly minimising the WLS cost function. This proves valuable in the deriving afresh of various existing and modified schemes, and helps us to show that the renormalisation approaches of Kanatani do not theoretically act to minimise the WLS cost function. A portion of this work serves to rationalise renormalisation, and several new variations on the theme are proposed. Various minimisation schemes are then tested. Experiments are carried out on the benchmark conic fitting problem of estimating ellipses from synthetic data points and their covariances. When the data exhibit noise that is anisotropic and inhomogeneous, those methods that make use of covariance information perform markedly better than more traditional methods that do not. None of the methods outperforms the fundamental scheme. Thus, being in addition simply expressed and constituting a genuine minimiser of the WLS cost function, the fundamental scheme offers strong advantages over the alternatives considered. 


An efficient recursive factorization method for determining structure from motion

Y. Li, M.J. Brooks
International Conference on Computer Vision and Pattern Recognition - CVPR'99,
IEEE Computer Society, Fort Collins, Colorado, June 1999.
PDF (cvpr99.pdf, 97Kb)

A recursive method is presented for recovering 3D object shape and camera motion under orthography from an extended sequence of video images. This may be viewed as a natural extension of both the original (Tomasi and Kanade 1994) and the sequential (Morita and Kanade 1997) factorization methods. A critical aspect of these factorization approaches is the estimation of the so-called shape space (Morita and Kanade 1997), and they may in part be characterised by the manner in which this subspace is computed. If P points are tracked through F frames, the proposed recursive least-squares method updates the shape space with complexity O(P) per frame. In contrast, the sequential factorization method updates the shape space with complexity O(P^2) per frame. The original factorization methoqd is intended to be used in batch mode using points tracked across all available frames. It effectively computes the shape space with complexity O(FP^2) after F frames. Unlike other methods, the recursive approach does not require the estimation or updating of a large measurement or covariance matrix. Experiments with real and synthetic image sequences confirm the recursive method's low computational complexity and accuracy, and indicate that it is well suited to real-time applications.


Towards robust metric reconstruction via a dynamic uncalibrated stereo head

M.J. Brooks, L. de Agapito, D.Q. Huynh, L. Baumela
Image and Vision Computing, 16, 14, Dec. 1998, pp. 989-1002.
PDF (ivc.pdf, 0.39Mb)

We consider the problem of metrically reconstructing a scene viewed by a moving stereo head. The head comprises two cameras with coplanar optical axes arranged on a lateral rig, each camera being free to vary its angle of vergence. Under various constraints, we derive novel explicit forms for the epipolar equation, and show that a static stereo head constitutes a degenerate camera configuration for carrying out self-calibration. The situation is retrieved by consideration of a stereo head undergoing ground plane motion, and new closed-form solutions for self-calibration are derived. An error analysis reveals that reconstruction is adversely affected by inward-facing camera vergence angles that are similar in value, and by a principal point location whose horizontal component is in error. It is also shown that the adoption of domain-specific robust techniques for computation of the fundamental matrix can significantly improve the quality of scene reconstruction. Experiments conducted with dynamic stereo head images confirm that avoidance of near-degenerate configurations and use of robustness techniques are essential if reliable reconstructions are in future to be attained.


Robust determination of structure from motion in the uncalibrated case

M.J. Brooks, W. Chojnacki, A. van den Hengel, L. Baumela
Proc. Fifth European Conference on Computer Vision - ECCV'98,
Freiburg, Germany, June 1998, Lecture Notes in Computer Science (Vol. 1), 1406, Springer Verlag, pp. 281-295.
PDF (eccv98.pdf, 0.14Mb)

Robust techniques are developed for determining structure from motion in the uncalibrated case. The structure recovery is based on previous work of the authors in which it was shown that a camera undergoing unknown motion and having an unknown, and possibly varying, focal length can be self-calibrated via closed-form expressions in the entries of two matrices derivable from an instantaneous optical flow field. Critical to the recovery process is the obtaining of accurate numerical estimates, up to a scalar factor, of these matrices in the presence of noisy optical flow data. We present techniques for the determination of these matrices via least-squares methods, and also a way of enforcing a dependency constraint that is imposed on these matrices. A method for eliminating outlying flow vectors is also given. Results of experiments with real-image sequences are presented that suggest that the approach holds promise.


Detecting suspicious background changes in video surveillance of busy scenes

D. Gibbins, G. Newsam, M.J. Brooks
Third IEEE Workshop on Applications of Computer Vision,
December 1996, Sarasota, Florida, USA, pp.22-26. PDF (wacv96.pdf, 0.14Mb)

Detecting background changes in scenes containing significant numbers of moving objects has several applications in video surveillance. One important example is the detection of suspicious packages left in busy airport terminals or train stations. This paper outlines a statistical approach to automatically detecting long term changes to the stationary component of a scene, and describes a prototype system which has been used to successfully demonstrate the feasibility of this approach.


Egomotion from optical flow with an uncalibrated camera

M.J. Brooks, L. Baumela, W. Chojnacki
Journal Optical Society of America A, 14, 10, Oct. 1997, pp. 2670-2677.
PDF (josa.pdf, 550Kb)
See also Technical Report (.pdf) (tr04.pdf, 36Kb)

In this paper, we analyse the motion of a camera having free intrinsic parameters. We define a free parameter to be one that is unknown and may vary continuously. A time-dependent epipolar equation is presented, followed by a formal definition of the time-derivative of the fundamental matrix for the case of a mobile camera. Next, differential forms of the epipolar equation are obtained. This may be seen as a recasting of the recent work of Vieville and Faugeras into an analytical framework. Critical to the approach is the determination, to within a common scalar factor, of two special matrices from optical flow data. The case of a camera with free focal length undergoing arbitrary motion is then considered in detail. Closed-form expressions are given, in terms of the entries of the two matrices, for the ego-motion parameters, as well as the focal length and its derivative.


Can the sun's direction be estimated from an image prior to the computation of object shape?

W. Chojnacki, M.J. Brooks, D. Gibbins
J. Mathematical Imaging and Vision, 7, 2, 1997, 139-147.

Various computational techniques have been developed that perform reasonably well in inferring shape from shading. However, these techniques typically require substantial prerequisite information if they are to evolve an estimate of surface shape. It is therefore interesting to consider how depth might be inferred from shading information without prior knowledge of various scene conditions. One approach has been to undertake a pre-processing step of estimating the light-source direction, thereby providing input to the computation of shape from shading. In this paper, we present evidence that a versatile light-source-direction estimator is unattainable, and propose that, in the absence of domain-specific knowledge, shape and light-source direction should be determined in a coupled manner.


Revisiting Pentland's estimator of light source direction

W. Chojnacki, M. J. Brooks, D. Gibbins
Journal of the Optical Society of America A, 11, 1994, 118-124.

This paper examines the pioneering method of Pentland for automatically estimating the direction of the ``sun'' from a single image. It is shown that, under the assumptions used in the derivation of the method, the estimate of source direction is erroneous. Specifically, it is shown that an image-based expression used in calculating source direction diverges to infinity as the density of image points is increased, and that the formula involving this expression is therefore incorrect. When the method is implemented, the flaw manifests itself in the undesirable dependence of the estimator upon image resolution. Supporting experimental evidence is given for this. An alternative source-direction estimator is proposed which is free of these drawbacks.


Maintained by Mike Brooks (last updated 05/02)