Image Analysis
Palm recognition for
secure identification (see the publication in Pattern Recognition journal
below,
LASR picture enhanced and shaded of child at 3 points in time to measure growth differences (supported by EPSRC grant) (Morris et al, 1999)
With the development of modern computers the collection of images has become a straightforward and powerful tool for recording information. However, the interpretation of images in terms of their constituent parts requires sophisticated computational and statistical models. For example, consider an image containing a hand.
In order to make objective use of this information, the computer needs to extract the outline of the hand from the image, to recognize it as a hand, and perhaps to compare it to other hands to see how it differs.
The effects of time can be included in image analysis. We can consider how an image changes over time, such as the changing appearance of a hand or face as an individual grows.
Some key contributions: fuzzy classification methods, fusion methods, multiple objects, and warping methods.
The main areas of application are
(i) medical imaging (ii) security and (iii) object recognition. The
methodologies developed in the papers below include
· An integrated Bayesian system for multiple object recognition under occlusion
· An algorithm for fuzzy classification using innovative statistical principle
· A fast and robust algorithm median filtering and mean thresholding (MFMT) for segmentation
· Bayesian hierarchical techniques for segmenting multiple modalities such as MR images under different resolutions
· A novel way of deforming images using prior information under a spline and kriging framework, applied to the description of shape changes, averaging of images, fusing of X-ray and NM images, tracking tagged MR images
· A reliable method for the automatic calculation of directional dermatoglyphic features using a semi-variogram, coupled with iterative conditional modes
· Shape analysis methods based on invariant distances for facial identification from photographs
· A new technology of ridge curves for laser scans of faces for describing shape before and after surgery
· Growth curves in shape spaces from laser scans
· Penalized likelihood approach to image warping and averaging
Read Paper:
2001 Glasbey, C. A. and Mardia, K. V. A penalised likelihood approach to image warping. J. Roy. Statist. Soc. B 63, 465-514.
Image warping is an essential tool used in image analysis and machine vision to assess severity of deformation in order to facilitate discrimination between objects and develop prototypes. This joint paper with Professor Chris Glasbey changed the traditional approach that has been in use over the last two decades by developing and synthesizing statistical techniques. The paper proposed novel concepts such as a penalized likelihood approach with the von Mises-Fourier similarity measure and the null distortion criterion.
Edited volumes
1989 Statistical Methods in Image Processing. A special issue for the J. Appl. Statist 16, no. 2.
1993 Image warping and Bayesian restoration with grey-level templates (with T. Hainsworth). In Statistics and Images I. Eds. K. V. Mardia and G.K. Kanji, Carfax Publishing, Abingdon, pp. 257-280.
1994 Statistics and Images: Vol.II. Carfax Publishing Co. Ltd., Abingdon, Oxfordshire.
1996 Image Fusion and Shape Variability Techniques, co-editors C. A. Gill & I. L. Dryden, Leeds University Press.
1997 The Art and Science of Bayesian Image Analysis, co-editors C. A.
Gill & R. G. Aykroyd.
1998 Medical Image Understanding and Analysis ’98, co-editors
Papers in Journals:
1984 Spatial discrimination and classification maps. Commun. Statist. A 13, 2181-2197. Corrections: Commun. Statist. A (1987), 16.
1988 Spatial classification using fuzzy membership models (with J. T. Kent). IEEE Trans. Pattern Anal. Machine Intelligence 10 659-671.
1988 A spatial thresholding method for image segmentation (with T. J. Hainsworth). IEEE Trans. Pattern Anal. Machine Intelligence 10, 919-927.
1988 A spatial thresholding method for image segmentation (with T. J. Hainsworth). IEEE Trans. Pattern Analysis and Machine Intelligence 10, 919-927.
1993 Techniques for on-line
gesture recognition on workstations (with N. Sheehy,
1996 Distributions of projective invariants and model-based machine vision (with C. R. Goodall, and A. N. Walder). Adv. Appl. Probab. 28, 641-661.
1996 Bayesian fused
classification of medical images (with M. A. Hurn, T. J. Hainsworth, J.
Kirkbride, and
1996 On statistical problems with face identification from photographs (with A. Coombes, J. Kirkbride, A. Linney and J.L. Bowie) J.Appl.Stat. 23, No.6, pp655-675.
1997 Bayesian image analysis. J. Theoretical Medicine 1, 63-77.
1997 Statistical methods for
automatic interpretation of digitally scanned fingerprints (with A. J.
Baczkowski, X. Feng, and T. J. Hainsworth). Special
Issue: Pattern Recognition Letters (E. S. Gelsema and L. N. Kanal, eds.) 18, 1197-1203. Elsevier Sci.,
1997 Deformable template
recognition of multiple occluded objects (with
1999 Stochastic templates for aquaculture images and a parallel pattern detector (with K. M. A. de Souza and J. T. Kent). J. Roy. Statist. Soc. Ser. C 48, 211-227.
2002 Matching of palm prints
(with
2006 Synthesis of image deformation strategies (with J. M. Angulo, and A. Goitia). Image and Vision Computing, 24, 1-12.
2006 Intrinsic random fields and image deformations” (with F.L. Bookstein, J. T. Kent, and C. R. Meyer). J. Mathematical Imaging and Vision, 26, 59-71.
2006 Penalised image averaging and discriminations with facial and fishery applications (with P. McDonnell and A. D. Linney). J. Appl. Statist. 33, 339-369.
Papers in Edited Volumes:
1995 Statistical approaches to image restoration (with J. D. Burrows, J. A. Little, C. C. Taylor and A. N. Walder). In ECOS ’95 European Convention on Security and Detection. IEEE Conf. Publ. 408, 59-63.
1999 Analysing growth in faces
(with R. J. Morris, J. T. Kent, M. Fidrich, R. G. Aykroyd, and A. Linney). In Proc. Conf. Imaging Science, Systems and
Technology,
2001 Shape in images. In:
Pattern Recognition from Classical to Modern Approaches. Editors Sankar K Pal
& Amita Pal, Indian Statistical Institute,
2003 Matching unlabelled configurations using the EM algorithm (C. C. Taylor, K. V. Mardia, and J. T. Kent). LASR2003 Proceedings. (Editors: R. G. Aykroyd, K. V. Mardia and M. J. Langdon). Leeds University Press, pp19-21.