Proceedinsg of the Second International workshop on Generative-Model Based Vision

Publication: Research - peer-reviewBook – Annual report year: 2004

View graph of relations

In the last decade, there has been a convergence of statistical and model-based approaches to computational vision. This is an ongoing process, leading to the emerging paradigm of generative-model-based (GMB) vision. This workshop/special issue aims to bring together researchers working on different problems within computational vision, who are interested in this paradigm. For the purposes of the workshop/special issue, GMB vision is a methodology which prescribes * the formulation of a parameterized probabilistic model of image generation; * estimation and/or maximization of the posterior probability (given an image or image sequence) of model parameters (state variables). Often, the generative model is used not only by the software developer in the formulation of the algorithm, but also by the algorithm itself as a component of an iterative estimation process. The state variables are whatever people want to know, (e.g. position, size, shape, color) about objects of interest. This definition is not meant to be dogmatic or to inhibit the development of the field, but only to give a focus to the presentations. In addition to papers describing new GMB algorithms, also appropriate to the workshop/special issue are * papers which focus on a detailed study of generative models (e.g. as models of the statistics of natural images); * papers which present new estimation methods for model parameters, or compare different estimation methods applied to the same generative model; * papers providing a GMB interpretation (or modification) of established vision algorithms. Examples of topics relevant to the workshop/special issue include, but are by no means limited to, the topics covered in the first GMBV workshop.
Original languageEnglish
PublisherIEEE Computer Society Press
StatePublished - 2004
Download as:
Download as PDF
Select render style:
Download as HTML
Select render style:
Download as Word
Select render style:

ID: 2710561