Thesis
Efficient bit-plane coding of lattice vector quantization using one-norm enumeration
Washington State University
Master of Science (MS), Washington State University
2013
Handle:
https://hdl.handle.net/2376/103142
Abstract
The sequence of Barnes-Wall lattices and their principal sublattices are a closely related family of 2n -dimensional binary lattices which can be constructed using coding formulas. Lattices are used for several applications including lattice vector quantization (LVQ) where the lattice structure can provide a significant portion of the possible granular gain. The given lattice codevector can be enumerated using the `1-norm. Representing the lattice vector in sign-magnitude form allows for encoding the lattice codevectors in a bit-stream that can be progressively decoded, bit-plane by bit-plane. Only the least significant bit-planes are lattice defining and should be encoded/decoded based on the lattice structure to achieve maximum coding compression. An encoding algorithm is considered that first encodes the one-norm, then the weights of the bit-planes, and finally the bit-planes conditioned on the weights. The algorithm is motivated by the properties of the conditional entropy of the lattice codevector. Adaptive binary arithmetic codes are used at each step of the encoding. The structure of the lattices impact the method of encoding the lattice defining bit-planes. Several sources are encoded in order to evaluate the performance of the one-norm encoding algorithm, including memoryless sources, audio, and image. The results are summarized as encoding rate vs. signal-to-noise ratio (SNR) for various sources and lattices.
Metrics
4 File views/ downloads
8 Record Views
Details
- Title
- Efficient bit-plane coding of lattice vector quantization using one-norm enumeration
- Creators
- Curtis Maddex
- Contributors
- Thomas R. Fischer (Degree Supervisor)
- Awarding Institution
- Washington State University
- Academic Unit
- Electrical Engineering and Computer Science, School of
- Theses and Dissertations
- Master of Science (MS), Washington State University
- Publisher
- Washington State University; [Pullman, Washington] :
- Identifiers
- 99900525064601842
- Language
- English
- Resource Type
- Thesis