lloyds

Optimize quantization parameters using Lloyd algorithm

collapse all in page

Syntax

[partition,codebook] = lloyds(training_set,initcodebook)
[partition,codebook] = lloyds(training_set,len)
[partition,codebook] = lloyds(___,tol)
[partition,codebook,distor] = lloyds(___)
[partition,codebook,distor,reldistor] = lloyds(___)

Description

[partition,codebook] = lloyds(training_set,initcodebook) optimizes the scalar quantization parameters partition and codebook for the training data in the vector training_set. initcodebook is the initial guess of the codebook values.

[partition,codebook] = lloyds(training_set,len) is the same as the first syntax, except that the scalar argument len indicates the size of the vector codebook. This syntax does not include an initial codebook guess.

example

[partition,codebook] = lloyds(___,tol) specifies the relative distortion tolerance,tol, in addition to any input argument combination in the previous syntaxes.

[partition,codebook,distor] = lloyds(___) additionally returns the final mean square distortion, distor.

[partition,codebook,distor,reldistor] = lloyds(___) additionally returns a value that depends on how the algorithm terminates, reldistor.

Examples

collapse all

Open Live Script

Create a sinusoidal signal to use as training data.

t = 0:1000;
training_set = sin(t*pi/500);

Set a codeword length of eight. This length corresponds to transmission via a 3-bit channel.

len = 8;

Optimize the quantization parameters. Use the optimized parameters to quantize a different sinusoidal signal. Plot the signal and the quantized signal.

[partition,codebook] = lloyds(training_set,len);
signal = cos(t*pi/250);
[index,quants] = quantiz(signal,partition,codebook);
plot(t,signal,t,quants,'.')

Figure contains an axes object. The axes object contains 2 objects of type line. One or more of the lines displays its values using only markers

Input Arguments

collapse all

Training data, specified as a vector.

Note

The function optimizes for the data in training_set. For best results, training_set should be similar to the data that you plan to quantize.

Data Types: double | single

Initial codebook guess, specified as vector. The length of the vector must be at least 2. Specify this input to provide an initial value for the iterative process described in Algorithms.

Data Types: double | single

Initial codebook length, specified as a positive integer. Specify this input when you want the function to generate an initial value for the iterative process described in Algorithms.

Data Types: double | single

Relative distortion tolerance, specified as a scalar. The Lloyd algorithm terminates once the relative change in distortion between iterations is less than this value.

Data Types: double | single

Output Arguments

collapse all

Distinct endpoints of different ranges, returned as a row vector. This output defines several contiguous, nonoverlapping ranges of values within the set of real numbers. The length of this vector is one less than the length of codebook. The data type is the same as that of training_set.

Quantization value for each partition, returned as a row vector. This output prescribes a common value for each partition in a scalar quantization. The length of this vector is equal to the length of initcodebook or the value of len, depending on which input you specify. The data type is the same as that of training_set.

Final mean square distortion, returned as a positive scalar. The function calculates the mean square distortion at each step of the Lloyd algorithm.

Final relative change in mean square distortion, returned as a positive scalar. If the algorithm terminates because of the first condition in Algorithms, reldistor is the relative change in distortion between the last two iterations. For the second condition, reldistor is equal to distor.

Algorithms

The lloyds function uses an iterative process to minimize the mean square distortion. Optimization processing ends when either:

  • The relative change in distortion between iterations is less than the value of tol.

  • The distortion is less than eps*max(training_set), where eps is the MATLAB® floating-point relative accuracy.

References

[1] Lloyd, S.P., “Least Squares Quantization in PCM,” IEEE Transactions on Information Theory, Vol. IT-28, March, 1982, pp. 129–137.

[2] Max, J., “Quantizing for Minimum Distortion,” IRE Transactions on Information Theory, Vol. IT-6, March, 1960, pp. 7–12.

Version History

Introduced before R2006a

See Also

Functions

Topics