This example shows the Vector Quantizer Design process using Generalized Lloyd Algorithm (GLA) for a two dimensional input.
|On this page…|
This example uses the Generalized Lloyd Algorithm, which is a common codebook design algorithm for digital image compression.
The two-dimensional codebook and corresponding Voronoi cells are shown in the upper plot. Individual codebook values are represented by small round symbols. The training set is shown in the plot as a light green cloud of points behind the cells. You can examine each codeword by right-clicking on it and holding the mouse button down. You can change codeword values by left-clicking on the symbol, holding the mouse button down, and dragging it to a new location. The Voronoi cells change to indicate the updated partition.
The lower plot shows the number of training vectors belonging to each Voronoi cell.
To run the example, you need to specify a training set and the number of levels. Enter the training set using the Training Set (TS) parameter. Enter an initial codebook in one of three ways:
Auto-generate - The example will select a set of initial codebook values based on the value entered in the Number of levels parameter.
User defined - Use the Initial codebook parameter to enter the codebook.
From plot - The codebook values shown in the plot are used.
The stopping criteria are "Relative threshold", "Maximum iteration", and "Whatever comes first". The Relative threshold and Maximum Iteration parameters are used to complete the stopping criteria.
Try varying one or more of the available parameters, run the example, and observe the change in the codebook after each iteration until the stopping criteria is met.
Gersho, A. and R. Gray. Vector Quantization and Signal Compression. Boston: Kluwer Academic Publishers, 1992.