Quantized Overcomplete Expansions: Analysis, Synthesis and Algorithms
Abstract
Linear transforms and expansions are the fundamental mathematical tools of signal processing. Yet the properties of linear expansions in the presence of coefficient quantization are not yet fully understood. These properties are most interesting when signal representations are with respect to redundant, or overcomplete, sets of vectors. Exploring the of quantization in overcomplete linear expansions is the unifying theme of this work. Two classes of overcomplete expansions were considered: fixed (frame) expansions and expansions that are adapted to the particular source sample, as given by matching pursuit.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1995
- Accession Number
- ADA637169
Entities
People
- Vivek K Goyal
Organizations
- University of California, Berkeley