Copy Link
Add to Bookmark
Report
VISION-LIST Digest Volume 12 Issue 38
VISION-LIST Digest Tue Aug 24 09:51:05 PDT 93 Volume 12 : Issue 38
- ***** The Vision List has changed hosts to TELEOS.COM *****
- Send submissions to Vision-List@TELEOS.COM
- Vision List Digest available via COMP.AI.VISION newsgroup
- If you don't have access to COMP.AI.VISION, request list
membership to Vision-List-Request@TELEOS.COM
- Access Vision List Archives via anonymous ftp to FTP.TELEOS.COM
Today's Topics:
Re: modelling image processing techniques
Optical filter spectral transmittance curves
Re: Machine Learning
Research position on foveated vision
Videometrics II session cha
BMVC93 - Final Programme and registration
----------------------------------------------------------------------
Date: 20 Aug 1993 18:38:07 GMT
From: cthang@focus.csl.uiuc.edu (Thang Cao Nguyen)
Organization: Center for Reliable and High-Performance Computing, University of Illinois at Urbana-Champaign
Subject: Re: modelling image processing techniques
Vision-List@TELEOS.COM writes:
> MODELLING THE PROPERTIES OF IMAGE SEGMENTATION TECHNIQUES
>As part of my PhD on crainial image segmentation I am interested in
>modelling the properties of various image segmentation techniques (eg
>edge detectors, region growers etc) in order to be able to reason or
>select the best segmentation strategy for detecting a particular
>anatomical structure within an image given that some of the properties
>of that structure and those surrounding it can be predicted from an
>anatomical model.
>Unfortunately I have been unable to discover many publications which
>discuss the problems of modelling the properties of segmentation
>techniques either by mathematical equations of heuristic rules. If
>anybody knows of any research which has been done into this topic I
>would be very grateful if they could email me with more details.
Happen to see an interesting paper on the characterization of
segmentation algorithms, despite being somewhat dated:
Kenneth I. Laws, On the Evaluation of Scene Analysis Algorithms,
Proc. Image Understanding Workshop, 1983.
The author is associated with SRI International. There are references to
some technical notes.
I myself would like to hear from more of this topics, especially those more
recent pulications.
Thang
------------------------------
Date: Sat, 21 Aug 93 06:58:50 -0500
From: mwolski@ecn.purdue.edu (Mark J Wolski)
Subject: optical filter spectral transmittance curves
I need to get a computer database of the spectral transmittance
curves of Schott and Hoya color glass filters (400-700nm at 10nm intervals).
Does anyone have access to such a file that they would be willing to
share ? (It would save me a lot of labor trying to extract the data
from catalogs.)
Thanks in advance.
Mark Wolski mwolski@ecn.purdue.edu
------------------------------
Date: Mon, 23 Aug 1993 10:20:32 +0200
From: Regis Vincent <Regis.Vincent@sophia.inria.fr>
Subject: Re: Machine Learning
To answer to your request. Here is a more complete article :
Machine Learning in Computer Vision : Similar Research Subject
I'm starting to work on my PHD thesis and i'm looking for people with a similar
research subject for possible collaboration.
My job is to connect a learning system to an E.S. shell for program
supervision in order to learn first the numerical values of numerical
parameters of the programs. The framework is the following :
A system for the supervision of numerical programs has to select, adapt
and schedule the most appropriate programs from a library to achieve a goal,
depending on the current situation. The reception of a request as input, produces as output
a sequence of commands to the shell.
The objective of supervision of programs in general is to facilitate the
automation of (or even to automate completely) an existing processing
activity, independently of any application. For example the application maybe
to extract some shape in an image and the supervision system has to trigger
the corresponding sequence of image processing program.
This task can be decomposed into two sub-tasks:
to select and schedule the vision modules, which corresponds to a planning
process, and to compute the parameter values, to run the programs
and to control their results, which corresponds to a control of execution
process.
Planning is the process of finding an ordered list of actions to achieve
a goal w.r.t. a given state of the world (selection and adaptation of
programs).
Actions are for example vision tasks (not robotics actions) and the state of the world
cannot be exhaustively described. Planning is based on the description of the
actions (not on the state of the world), and because the knowledge exists at
several levels of abstraction, the planning process uses skeletal
plans with refinement mechanisms.
Control of execution consists in an optimisation of the chosen plan.
In this framework, my first job is to connect a learning system in order to
learn the numerical values of numerical parameters of the programs.
The numerical programs have input parameters. These parameters are of two
types; they are either input data (digitized images ...), or numerical
arguments, that influence the processing and hence the result of the
processing (threshold, image resolution, ...).
It is often very difficult for the expert to provide reliable values for this
parameters because he often knows either average values, or values for
particular cases. The initial knowledge base provided by the expert doesn't
contain enough knowledge to solve all the cases.
Depending on the input data, the processing could lead to failures
(unsatisfactory result). The system must then adjust the values of some
parameters until it obtains a correct result. But if a similar case arises
afterwards, the system would not have recorded the correct values. A means to
complete the knowledge base is learning of numerical values in case of
failures of the initial knowledge base.
As input, we have a example of readjustement of an incorrect parameter value and as
output we want to produce a rule to initialize properly this parameter value
for similar cases.
This kind of rule contains expressions with numerical values in both the premises and
the conclusions.
I'm trying to find a learning strategy that is able to generalize a given set
of example rules.
For example : if (a = 3) then (b=4)
if (a = 3.1) then (b=4.1)
Regis VINCENT
rvincent@sophia.inria.fr
------------------------------
Date: Tue, 24 Aug 93 9:15:03 METDST
From: Giulio Sandini <giulio@vision.dist.unige.it>
Subject: Research position on foveated vision
FOVEA: Call for Visiting Researchers
The Commission of the European Communities (CEC) has awarded
a Human Capital and Mobility fellowship grant to a
consortium of 3 research organisations on the topic of
"SPACE VARIANT VISION AND LOG-POLAR MAPPING.
The Project and the profile of the candidates are outlined
below.
OUTLINE OF THE RESEARCH AREA
The aim of this project is to create and disseminate
scientifc know-how in the use and development of a space-
variant CCD visual sensor whose characteristics are
derived from the distribution of the photoreceptors in the
human retina (retina-like sensor). The proposers in this
project have been among the originators of the project and
fabrication of the CCD sensor above mentioned.
The reseach project that the researchers will be
involved into is multidisciplinary and, as such, the
scientists working in the different labs, will have to
maintain strong links between them in order to achieve the
results we are aiming at. In fact a visual system based on
this particular kind of sensors requires the integration
of know-how in the fields of:
1) Computer Vision
2) Motor Control and Robotics
3) VLSI (in particular CCD) design and fabrication
More specifically the scientific problems that will be
investigated are:
1) Visual Recognition in the log-polar space (the image
sampled by this particular sensor is "topologically
distorted" by a, so-called, log-polar mapping);
2) Visuo-motor integration and computation of dynamic
features (e.g. optical flow, time-to-impact, tracking);
3) Implementation of gaze-control strategies on a
binocular head equipped with retinal sensors.
4) Visuo-motor self-calibration;
5) Study and development of on-chip processing
capabilities;
6) Study and development of novel processes for CCD
fabrication;
7) Investigation of potential industrial applications in
the fields of robotics and telecommunication.
Within these major themes the research activity will be
devoted both to practical implementations of vision
algorithms based on the retinal sensor and on the
theoretical aspects related to the design of special
purpose, low cost, high performance visual systems based on
special purpose sensors
GOALS OF THE PROJECT
The main results we intend to achive are:
1) Development of visual routines on a stereo visual system
based on two cameras equipped with the CCD space-variant
visual sensors and mounted on a 4 degrees of freedom robot
head.
2) Design of an improved version of the CCD sensor (both
resolution and on-chip processing power will be investigated
as well as new technological solutions);
3) Identification and preliminary evaluation of the
performance of the sensor in industrial applications.
PARTICIPATING LABORATORIES
Three laboratories participate in the project:
DIST, University of Genova
LIRA-Lab
Via Opera Pia 11a
16145 GENOVA - Italy
Research Interest:
Robot Visin, Gaze Control, Applications of Space-Variant
Vision
Duration of fellowship: 6 months
Contact Person: Prof. Giulio Sandini
Phone: +39 10 3532779
Fax: +39 10 3532948
e-mail: giulio@dist.dist.unige.it
*****
Scuola Superiore S. Anna
ARTS Lab
Via Carducci, 40
57127 Pisa, Italy
Research interests:
orientation (visual servoing), sensorimotor integration,
mecatronics
Duration of fellowship: 6 months
Contact Person:Prof. Paolo Dario
Phone: +39-50-559207/560108
Fax: +39-50-559215
email: dario@arts.sssup.it
*****
IMEC (Interuniversitary Microelectronic Center)
vzw Kapeldreef 75
Leuven B-3030
Belgium
Research Interest: VLSI and CCD design and fabrication
Duration of fellowship: 9 months
Contact Person: Dr. Lou Hermans
Phone: +32 16 281211
Fax: +32 16 22.94.00
e-mail: hermans@imec.be
PROFILE OF CANDIDATES
The candidates must have at least two years experience of
work in research environment and must be in a pre-doctoral
condition.
In order to facilitate the multidisciplinary approach of the
project, the two periods offered by DIST and IMEC could be
assigned to the same person with the idea that the
fellowship holder will work for 6 months at DIST and for the
following 9 months at IMEC. The reason being the fact that
it is our intention to design a new version of the sensor
incorporating some of the requirements derived by the need
of visual processing. The period at DIST will give the
candidate the necessary background in term of visual
processing which will be used, during the following period
at IMEC, to design a new sensor. The position offered by
SSSA for a duration of 6 months will be devoted, mainly, to
work on the design of an improved version of the CCD camera.
Therefore the ideal candidates should have the following
expertize:
DIST-IMEC candidate: Visual processing and/or VLSI design.
SSSA candidate: Electronic design.
According to the restrictions of the HCM program the DIST-
IMEC candidate must not have an Italian or a Belgian
citizenship and the SSSA candidate must not have an Italian
citizenship. Both must have a citizenship in a Country of
the CEC.
The CEC funds will pay for salary, travel, overheads to the
host institution and all expenses during the visit.
If you or one of your colleagues/staff are interested in
applying for a research visit, please look at the list of
organisations and identify one whose research interests are
compatible with yours. You should then make contact with the
designated person in the CHOSEN ORGANIZATION DIRECTLY.
If you need more information about the general topics of the
project you can contact me at my e-mail address.
Giulio Sandini
------------------------------
Date: Mon, 23 Aug 1993 14:43:36 +1000
From: Mark Shortis <mark_shortis@muwayf.unimelb.edu.au>
Subject: Videometrics II session cha
Time:2:41 PM
OFFICE MEMO Videometrics II session chairs
Date:8/23/93
As an extra piece of information regarding the Videometrics II
conference to be held in Boston during September 9-10, session 4 will
be chaired by Kam Wong (USA) and session 5 will be chaired by John
Trinder (Australia).
Regards,
Dr. Mark R. Shortis, Mark_Shortis@mac.unimelb.edu.au
Senior Lecturer,
Dept. of Surveying and Land Information,
University of Melbourne, Telephone +613 344 6401
Parkville 3052, AUSTRALIA. Facsimile +613 347 2916
"An expert is a person who has made all possible errors in a very
narrow field of endeavour"
------------------------------
Date: Mon, 23 Aug 1993 09:44:45 +0100 (BST)
From: J.Illingworth@ee.surrey.ac.uk
Subject: BMVC93 - Final Programme and registration
FINAL PROGRAMME
*******************************************
* LAST CHANCE TO REGISTER FOR BMVC93 *
*******************************************
BOOK YOUR PLACE NOW TO AVOID DISAPPOINTMENT
------------------------------------------
British Machine Vision Conference BMVC'93
------------------------------------------
PROVISIONAL PROGRAMME AND REGISTRATION FORM
*******************************************
University of Surrey, Guildford. U.K.
21-23 September 1993
The British Machine Vision Conference is the main annual U.K. forum for
the presentation of research results in the areas of machine vision and
pattern recognition. This two and a half day conference will take place
from Tuesday 21 September to Thursday 23 September on the pleaseant
landscaped campus of the University of Surrey, Guildford.
The conference will be accompanied by a small exhibition of commercial
and industrial products related to computer vision.
The conference will also be preceeded by a half-day topical seminar on
``Wavelets'' which will be given by Roland Wilson of Warwick University.
The University of Surrey is situated in the historic cathedral town of
Guildford. It is only 30 miles from Central London and has excellent
road and rail links to the rest of the U.K.. It is also sited close to
and halfway between Londons major international airports, Heathrow
and Gatwick.
Monday 20 September 1993
Afternoon Tutorial: 2 p.m. to 5:30 p.m.
AN INTRODUCTION TO WAVELET TRANSFORMS
R. Wilson, Department of Computer Science , Warwick University.
The last five to ten years have seen the development of a new class of
signal representation - the wavelet transform - which has placed
multiresolution signal analysis on a rigorous footing and has opened up
new possibilities for efficient multiresolution methods of image coding
and analysis. While there have been many publications on wavelets, a
combination of some unusual notation and sophisticated mathematics have
not helped the general reader to assess their potential in applications.
The aim of the tutorial will be to answer the following questions :
* What are wavelets and how do they relate to more established
multiresolution methods ?
* How can wavelet transforms be computed efficiently ?
* What can they be used for ?
* What are their limitations ?
The only prerequisite assumed will be a degree of familiarity with the
conventional signal processing concepts of filtering and linear
transformation. The tutorial session will include examples showing how
various wavelet transforms may be used in low-level vision tasks,
such as boundary detection, stereopsis and texture analysis.
It will be concluded with a survey of recent developments in the field.
\medskip
Cost: 20 pounds (free to full-time research students)
Tuesday 21 September 1993
Invited Lecture
***************
``Issues in Robot Vision''
Prof Goesta Granlund, University of Linkoeping, Sweden
Pattern Analysis
****************
``Automatic Machine Learning of Decision Rules for Classification
Problems in Image Analysis''
Pudil, Novovicova and Kittler (Univ of Surrey)
``Testing Face Recognition''
Robertson and Craw (Aberdeen Univ)
Active Vision
*************
``Saccade and Pursuit on an active Head/Eye Platform''
Bradshaw, McLauchlan, Reid and Murray (Oxford Univ)
``Estimating Rigid 3D Motion by Stereo Fixation of Vertices''
Yang and Illingworth (Univ of Surrey)
``Active Fixation for Scene Exploration''
Brunnstrom, Eklundh and Uhlin (KTH, Stockholm)
``Active Animate Stereo Vision''
Urquhart, Siebert, McDonald and Fryer (Turing Inst. and Strathclyde Univ)
``Model Driven Selection using Texture''
Syeda-Mahmood (MIT)
Recognition and Pose
********************
``Recognising Objects on the Ground Plane''
Tan, Sullivan and Baker (Reading Univ)
``Determination of the pose of an articulated object from a single
perspective view''
Dhome, Yassine and Lavest (Blaise Pascal Univ, France)
``Statistical Partial Constraints for 3D Model Matching and Pose
Estimation Problems''
Waite, Orr, Fisher and Hallam (Edinburgh Univ and ARR Salford)
``Relative positioning from Model Indexing''
Carlsson (KTH, Stockholm)
``On Computing the perspective transformation matrix and camera
parameters''
Tan, Sullivan and Baker (Reading Univ)
Shape 1
*******
``Isotropic Regularisation''
Nielsen (DIKU, Copenhagen)
``Robust Shape from Shading''
Jones and Taylor (Manchester Univ)
``Model Construction from a single perspective view using shape from
symmetry''
Attwood, Sullivan and Baker (Reading Univ)
``Shape from Copies''
Van Diest, Moons, Van Gool and Oosterlinck (Leuven Univ)
Wednesday 22 September 1993
Invited Lecture
***************
``Geon based Object Recognition''
Prof Irving Biederman, University of California
Texture
*******
``Parallel Texture Region Segmentation using a Pearl Bayes Network''
Ducksbury (DRA, Malvern)
``Neural networks for the Texture Classification of Segmented Regions of
Forward Looking Infrared Images''
Haddon and Boyce (DRA, Farnborough and Kings College, London)
Motion and Tracking
*******************
``Strategies for Tracking Tokens in a cluttered scene''
Zhang (INRIA, France)
``Perspective alignment Back-projection for Monocular Tracking of
Solid Objects''
Verghese (Univ of Toronto, Canada)
``Bayesian Net for Mapping Contextual Knowledge to Computational
Constraints in Motion Segmentation and Tracking''
Gong and Buxton (QMW, London and Sussex Univ)
``Estimation of Complex Multimodal Motion: an approach based on
Robust Statistics and Hough Transform''
Bober and Kittler (Univ of Surrey)
``Modelled Object Pose Estimation and Tracking by Monocular Vision''
Daucher, Dhome, Lapreste and Rives (Blaise Pascal Univ, France)
Segmentation
************
``The Coherent Circle Hough Transform''
Atherton and Kerbyson (Warwick Univ)
``Computationally efficient Hough Transform for 2D object location''
Davies (RHBNC, London)
``Segmentation of 3D Articulated Objects by Dynamic Grouping of
Discontinuities''
Borges and Fisher (Edinburgh Univ)
``Multiscale Hierarchical segmentation''
Griffin, Robinson and Colchester (Guys Hospital)
``Edge Enhamcement and fine feature restoration of segmented objects
using pyramid based adaptive filtering''
Grace and Spann (Birmingham Univ)
Shape 2
*******
``Adding Gray Level Information to Point Distribution Models
using Finite Elements''
Marchant (AFRC, Silsoe)
``A Generic System for Classifying Variable Objects using
Flexible Template Matching''
Lanitis, Taylor and Cootes (Manchester Univ)
``Model Based Interpretation of 3D medical images''
Hill, Thornham and Taylor (Manchester Univ)
``A Distributed Approach to Image Imterpretation using Model
Based Spatial Reasoning''
Ratter, Baujard and Taylor (Manchester Univ)
Thursday 23 September 1993
Affine Structure
****************
``Extracting Structure from Single Affine Views of 3D Point Sets with
One or two Bilateral symmetries''
Fawcett, Zisserman and Brady (Oxford Univ)
``Finding Point Correspondences in Motion Sequences Preserving
Affine Structure''
Sudhir, Banjeree and Zisserman (IIT, New Delhi and Oxford Univ)
``Towards 3D object model acquisition and recognition using
3D affine invariants''
Vinther and Cipolla (Cambridge Univ)
``Epipolar Estimation using Affine Motion-parallax''
Lawn and Cipolla (Cambridge Univ)
``Uncalibrated Stereo Hand-Eye Coordination''
Hollinghurst and Cipolla (Cambridge Univ)
Applications
************
``Face Segmentation for Identification using Hidden Markov Models''
Samaria (Cambridge Univ)
``Contextual Classification of Cracks''
Bryson, Dixon, Hunter and Taylor (Manchester Univ)
``Seismic Time Section Analysis using Machine Vision''
Tu, Zisserman and Mason (Oxford Univ)
``Use of Geometric Histograms for Model Based Object Recognition''
Evans, Thacker and Mayhew (Sheffield Univ)
``Automated registration of images of different anatomical structures
using knowledge of adjacency and training by Registered datasets''
Hill and Hawkes (Guys Hospital, London)
Posters
*******
``A Robust Real Time Face Location Algorithm for Videophones''
Ponticos (Philips Research Labs)
``Attentive Visual Tracking''
Roberts and Charnley (Southampton Univ)
``Colour and Texture in Cloud identification:
a comparison of Neural Net and Bayesian approach''
Richards and Sullivan (Reading Univ)
``Monoplanar Camera Calibration: iterative multistep approach''
Batista, Dias, Araujo and Traca de Almeida (Univ of Coimbra, Portugal)
``Computation of vehicular Trajectories using a Neural Network''
Zhang (UCL, London)
``Elastic Models and Self-Organising Maps for Chromosome Classification''
Turner, Austin, Allinson and Thompson (York Univ)
``Texture Analysis using artificial neural nets and mode filters''
Greenhill and Davies (RHBNC, London)
``Automated detection of breast asymmetries''
Miller and Astley (Manchester Univ)
``An Optimised Vanishing Point Detector''
Palmer and Tai (Surrey Univ)
``Generation, Verification and Localisation of Object Hypotheses
based on Colour''
Matas, Marik and Kittler (Surrey Univ)
``Occlusion Analysis of Spatiotemporal Images for Surface Reconstruction''
Yasuno and Suzuki (NTT, Japan)
``Advances in Model based Traffic Vision ''
Worrall, Sullivan and Baker (Reading Univ)
``Invariant Fitting of arbitary single-extremum surfaces''
Fitzgibbon and Fisher (Edinburgh Univ)
``Multi-scale Salience Distance Transforms''
Rosin and West (Curtin Univ, Australia)
``Hierarchical Matching beats the Non-wildcard and Interpretation Tree
Matching Algorithms''
Fisher (Edinburgh Univ)
`Graduated Non-Convexity by Smoothness Focusing''
Nielsen (DIKU, Copenhagen)
``Parameter free Stereo Matching Algorithm''
Ruihua, Thonnat and Berthod (INRIA, France)
``Visually Salient 3D Model Acquisition from Range Data''
Bispo, Fitzgibbon and Fisher (Edinburgh Univ)
``Bayesian Corner Detection''
Zhang and Haralick (Univ of Washington, USA)
``Active Shape Model Search using Local Grey Level Models:
a quantitative evaluation''
Cootes and Taylor (Manchester Univ)
``Motion Correspondence using a Neural Network''
Sarigiandis and Pycock (Birmingham Univ)
``Affine Stereo Calibration for Relative Stereo Reconstruction''
Quan (LIFIA, France)
REGISTRATION FORM
^^^^^^^^^^^^^^^^^
4th British Machine Vision Conference
University of Surrey, Guildford. 21-23 September 1993
Full Name:.......................................................
Affiliation:.....................................................
Address:.........................................................
.........................................................
.........................................................
Signature: ..............................................
Date: .................................
Tel:..........................
Fax:..........................
Email:........................
Sex M/F: ................ Vegetarian Y/N:.........................
The full conference BMVC'93 begins at 9~a.m. on 21 Sept and
ends at lunchtime on 23 Sept. The conference will be preceded
on the afternoon of 20 Sept by a tutorial on ``Wavelets'' given
by Dr Roland Wilson of Warwick University.
Throughout the conference there will be a small exhibition of commercial
and industrial products related to computer vision.
All registration fees include attendance at the conference, a copy of
the conference proceedings, attendance at the conference dinner on
Wednesday 22 Sept and lunches on 21, 22 and 23 Sept.
The residential fee also includes bed and breakfast accommodation on
the evenings of 20, 21 and 22 Sept.
To register please choose between the residential or non-residential
package and fill in the appropriate options and amounts to pay in the
table below.
\medskip
| Residential | Non-Residential | To Pay
BMVA member Conf. Registration | 230 | 180 |
fee | | |
Non-BMVA member Conf. | 250 | 200 |
Registration fee | | |
Late Registration surcharge | 20 | 20 |
(after 9 Aug 93) | | |
1994 BMVA Membership |standard | 15 | 15 |
Renewal fee |student | 7:50 | 7:50 |
Wavelets Tutorial |standard fee| 20 | 20 |
|student fee | free | free |
TOTAL | | |
Payment must be in pounds sterling via cheques
made payable to ``University of Surrey''. This form and payment should
be returned to:
Dr John Illingworth (BMVC93),
Department of Electronic and Electrical Engineering,
University of Surrey, Guildford, GU2 5XH.
United Kingdom.
tel: +44 483 509835
fax: +44 483 34139
e-mail: J.Illingworth@ee.surrey.ac.uk
A limited number of student bursaries will be available. For details
apply to:
Geoff Sullivan,
Dept of Computer Science,
University of Reading, Reading.
RG6 2AX.
Preference will be given to students presenting a paper.
Dr. J. Illingworth, | Phone: (0483) 509835
V.S.S.P. Group, | Fax : (0483) 34139
Dept of Electronic and Electrical Eng, | Email: J.Illingworth@ee.surrey.ac.uk
University of Surrey, |
Guildford, |
Surrey GU2 5XH |
United Kingdom |
------------------------------
End of VISION-LIST digest 12.38
************************
