Publications (3)0 Total impact
-
[show abstract]
[hide abstract]
ABSTRACT: This paper describes a new Fixed-rate Entropy-constrained Vector Quantization (FEVQ) scheme for stationary memoryless sources based on a sequential search procedure. It is shown that the proposed algorithm results in a substantial reduction in the complexity while the degradation in performance is negligible.
10/2002;
-
[show abstract]
[hide abstract]
ABSTRACT: This paper describes two new fixed-rate entropy-coded quantization methods for stationary memoryless sources where the structure of code-words are derived from a variable-length scalar quantizer. In the first method, we formulate the quantization as a zero-one integer optimization problem. We show that the resulting integer program can be closely approximated by solving a simple linear program. The result is a Lagrangian formulation which adjoin the constraint (length) to total distortion. Unlike the previous methods with a fixed Lagrangian multiplier (fixedslope, and variable rate output), we use an iterative algorithm to optimize Lagrangian function while updating the slope of the function until the cost constraint is satisfied with equality (ensure to be fixed-rate). In order to achieve some part of packing gain, we combine the process of trellis encoding with that of quantizer shaping using linear programming. This results in an iterative use of Viterbi algorithm for optimizing the Lagrangian function. For the important class of sources with a monotonically decreasing density, we present another fixed-rate method with negligible complexity. Numerical results show an excellent performance with a small complexity for the proposed schemes as compared to previously known methods.
11/2001;
-
[show abstract]
[hide abstract]
ABSTRACT: Variable-length codes (e.g. Huffman codes) are commonly employed to minimize the average binary codeword length in the quantization of discrete sources. However, the effect of error propagation limits the usefulness of variable length codes for the transmission over noisy channels. To avoid this shortcoming, one can use a fixed-rate entropy constrained quantizer, limiting the propagation of error within a single block. This paper presents the optimal decoder for a fixed-rate entropy constrained scalar quantizer. We model all the possible combinations of codewords in a given block by a trellis structure where the Viterbi algorithm is used to choose the most likely path through this trellis. Numerical results are presented indicating a substantial reduction in the end-to-end source distortion with respect to the conventional decoding methods.
11/2001;
