Multiscale Statistical Models for Signal and Image Processing
Abstract
We are developing a general theory for multi scale signal and image modeling, processing, and analysis that matched to singularity-rich data, such as transients and images with edges. Using a linguistic analogy, our model can be interpreted as grammars that constrain the wavelet vocabulary. Our investigation focuses on probabilistic graph models (tree-based hidden Markov models) that can accurately, realistically, and efficiently represent singularity structure in the wavelet domain. Grammar design is being guided by a detailed study of the final structure of singularities using Besov spaces and multifractal analysis.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 2004
- Accession Number
- ADA425177
Entities
People
- Richard G. Baraniuk
Organizations
- Rice University