Research in Modeling Patterns and Texture

How do humans quickly sort photos into categories, at only a glance? Or recognize a shot in spite of changed viewpoint? What computer measures can mimic human "perceptual similarity?" How do we weave context into such measures? What video/image models facilitate access to content while representing the signal as efficiently as (or more efficiently than) current coding methods? What simple (nonlinear?) models give the best control over manipulating picture content?

Even though one model may be capable of representing everything, no one model appears to be best at everything. The research challenge is to understand the strengths and weaknesses of promising models, and determine how these relate to other models. The long-term goal is to construct a "society of models" which works together, giving flexible and powerful performance for a variety of problems. This involves giving the computer the ability to learn when one model is best, and the ability to intelligently combine multiple models when that leads to an even better result.

This research to date has focused on models of image and video, especially those which represent collective low-level visual features such as color and texture ("vision texture").

Selected Publications

To browse or download these documents, visit our old tech-reports page.

Research overview:

``A Society of Models for Video and Image Libraries,'' R. W. Picard, IBM Systems Journal, Vol 35, Nos 3&4, 1996, pp. 292-312, TR #360.

Society of models & learning how to select/combine them:

``Interactive Learning using a Society of Models,'' T. P. Minka and R. W. Picard, Pattern Recognition, 30(4), 1997, TR #349.

``Vision Texture for Annotation,'' R. W. Picard and T. P. Minka, ACM/Springer Journal of Multimedia Systems, Vol. 3, pp. 3--14, TR #302.

Modeling different camera views & mosaicing images of different views:

``Video Orbits of the Projective Group: A New Perspective on Image Mosaicing," S. Mann and R. W. Picard, IEEE Trans. Image Processing, TR #338.

``Virtual Bellows: Constructing High Quality Stills from Video,'' S. Mann and R. W. Picard, Proc. ICIP, Austin, TX, Nov. 1994, TR #259.

Perceptually-motivated models for pattern retrieval:

``Periodicity, directionality, and randomness: Wold features for image modeling and retrieval,'' F. Liu and R. W. Picard, IEEE Trans. PAMI, Vol 18, No. 7, pp. 722-733, TR #320.

``A New Wold Ordering for Image Similarity,'' R. W. Picard and F. Liu, Proc. ICASSP , Adelaide, Australia, Vol. V, pp. 129-132, Apr 1994, TR #237.

Orientation-based features for perceptual and "quick glance" recognition:

``Texture Orientation for Sorting Photos `at a Glance,''' M. M. Gorkani and R. W. Picard , Proc. ICPR, Oct 1994, TR #292.

``Finding Perceptually Dominant Orientations in Natural Textures,'' R. W. Picard and M. M. Gorkani, Spatial Vision, Vol. 8, No. 2, pp. 221-253, 1994, TR #229.

Modeling high-dimensional probability distributions:

``Cluster-Based Probability Model and its Application to Image and Texture Processing,'' K. Popat and R. W. Picard IEEE Trans. Image Processing, TR #351.

``Novel Cluster-Based Probability Model for Texture Synthesis, Classification, and Compression,'' K. Popat and R. W. Picard Proc. of the SPIE Visual Comm. and Image Proc., Boston, MA, pp. 756-768, Nov 1993, TR #234.

Auto-regressive and random field models, theory and applications:

"Temporal Texture Modeling", M. Szummer and R. W. Picard, Proc. Int. Conf. Image Proc., Lausanne, Sep. 1996, TR #381.

``Gibbs Random Fields, Cooccurrences, and Texture Models,'' I. M. Elfadel and R. W. Picard, IEEE Trans. PAMI, Vol. 16, No. 1, pp. 24-37 Jan 1994, longer version in TR #204

``Random Field Texture Coding,'' R. W. Picard, Society for Information Display Int. Symposium Digest, Boston, MA, Vol. XXIII, pp. 685-688, May 1992, TR #185.

``Gibbs random fields: temperature and parameter analysis,'' R. W. Picard, ICASSP, San Francisco, CA, Vol. III, pp. 45-48, Mar 1992, TR #177.

``Structure of Aura and Co-occurrence Matrices for the Gibbs Texture Model,'' R. W. Picard and I. M. Elfadel, J. of Mathematical Imaging and Vision, No. 2, pp. 5-25, 1992, TR #160.

Reaction-diffusion models, morphogenesis, new M-Lattice:

``Color Halftoning with M-Lattice,'' A. Sherstinsky and R. W. Picard, Proc. ICIP, Washington DC, Oct. 1995, TR #336.

``M-Lattice: From Morphogenesis to Image Processing,'' A. Sherstinsky and R. W. Picard, IEEE Transactions on Image Processing, Vol. 5, No. 7, pp. 1137-1150, July 1996, TR #299.

Modeling flames in computer graphics:

``Synthesizing Flames and their Spreading,'' C. Perry and R. W. Picard, Proc. of the Fifth Eurographics Workshop on Animation and Simulation, Oslo, Norway, Sep 1994, TR #287.

OLD Vision & Modeling Group Home Page