glo.gif (1993 bytes)Glossary

 

1

13-segment law

A variant of the A-law encoding method. The continuous linear/logarithmic characteristic of the A-law is approximated with 13 linear segments. This method was developed for the standard PCM voice transmission in early systems. The voice signal was sampled at 12 bits per sample, and then downsampled to 8 bits using a simple translation table which translated each 12-bit sample to an 8-bit sample, according to the segmented characteristic.

A

A-law encoding/compression

A variant of the basic PCM encoding method, involving non-linear quantization. The quantization levels are distributed according to the logarithmic A-law. It has linear characteristic around zero and logarithmic (hence the name) characteristic for higher amplitudes. In the logarithmic part of the characteristic, the signal-to-noise ratio is constant and worse than with linear quantization. This is the price for having improved ratio for the low signal levels (in the linear part of the characteristic)
Related terms: u-law, 13-segment law
Related link: PCM: A-law compression

ADPCM: Adaptive Differential Pulse Code Modulation

A modification to the DPCM encoding scheme, which results in increased compression rate with an acceptable quality loss. The step size for the difference quantization is adapted to the momentary rate of change of the input signal. This method is used in audio signal compression.
Related terms: PCM, DPCM
Related link: DPCM: info on ADPCM

Arithmetic Coding - Adaptive

Arithmetic Coding - Static

Arithmetic Coding - Encoding Interval Range

C

Character

This term is used in the dictionary-based compression methods. It denotes the principal atomic data unit in the uncompressed data. A sequence of characters gets replaced with a single code word, according to a certain principle defined in the encoding algorithm.
Related term: charstream
Related links: LZ77: definition of terms, LZ78: definition of terms

Charstream

A term used in the LZ78-family of compression methods. It denotes the uncompressed sequence (stream) of characters.
Related term: codestream
Related link: LZ78: definition of terms

Codestream

A term used in the LZ78-family of compression methods. It denotes the encoded (compressed) data sequence.
Related term: charstream
Related link: LZ78: definition of terms

Code word

A general term used in data encoding and compression terminology. It denotes the basic, atomic binary data unit which is assigned a certain meaning. Its meaning depends on the encoding/compression scheme.
Related term: codestream
Related links: PCM: introduction, LZ78: definition of terms.

D

Dictionary

A term used in the LZ78-family of compression methods. The dictionary is an indexed list of phrases seen in the charstream. Whenever a phrase already stored in the dictionary reoccurrs in the charstream, it is encoded with its index number in the dictionary.
This term is also used in a more general meaning, denoting a generic container of phrases. The window in the LZ77-family can also be interpreted as a dictionary.
Related link: LZ78: definition of terms

Dictionary-based compression

This type of compression uses the propriety of many data types to contain repeating code sequences. These sequences are stored in a dictionary and each occurrence is replaced with the information about the position of the sequence in the dictionary. There are two families of dictionary-based methods: the LZ77-family and the LZ78-family.
Related terms: character, charstream, code word, codestream, dictionary, encoding step, lookahead buffer, phrase, pointer, prefix, string, window
Related link: Overview of dictionary methods

Digitizing (digitalization)

The process of recording an analog signal in a digital form. The signal is represented with a sequence of binary digits (bits). The basic method of signal digitizing is PCM.
Related terms: DPCM, quantization, sample
Related links: PCM, DPCM

DPCM: Differential Pulse Code Modulation

A modification to the PCM encoding scheme. Based upon the previously encountered samples, DPCM predicts the value for the next sample and records the difference between this value and the value that really occurred. This difference is usually small, so it can be encoded with less bits than the PCM sample. The predicted value is called the predictor.
Related terms: ADPCM, digitizing, predictive coding, quantization, sample
Related links: PCM, DPCM

E

Encoding step

A term used in the dictionary-based compression methods. The encoding algorithm completes one encoding step each time it produces output, thus completing the encoding of the current character sequence and moving on to the next.
Related links: LZ77: example, LZSS: example, LZ78: example, LZW: example

I

Image

Two-dimensional signal perceived by human visual system. A digital image is a two-dimensional array of picture elements called pixels.
Related links: CCITT fax compression, DPCM: example

L

Lookahead buffer

A term used in the LZ77-family of compression methods. It contains the data to be compressed. The encoding algorithm tries to find a match in the previous data (contained in the window) for the sequence at the beginning of the lookahead buffer.
Related link: LZ77: definition of terms

LZ77

A dictionary-based compression method, developed in 1977 by Lempel and Ziv. Replaces a re-occurring sequence with a pointer to its earlier occurrence.
Related term: LZSS
Related link: LZ77

LZ78

A dictionary-based compression method, developed in 1978 by Lempel and Ziv. It builds a dictionary of phrases seen in the data. When a phrase reoccurrs, it replaces it with an index number of the phrase in the dictionary.
Related term: LZW
Related link: LZ78

LZMW

A dictionary-based compression method. This is a variation of the LZW method. A new string in the dictionary is generated by concatenating the two previous ones.
Related link: LZW: info on LZMW

LZSS

A dictionary-based compression method, developed in 1982 by Storer and Szymanski. This is a variation of the LZ77 method.
Related link: LZSS

LZW

A dictionary-based compression method, developed in 1984 by Terry Welch. This is a variation of the LZ78 method.
Related link: LZW

M

mu-law

See u-law.

P

PCM: Pulse Code Modulation

The basic method of digitally recording (digitizing) an analog signal. Based on taking samples of the signal through time or space.
Related terms: DPCM, quantization
Related links: PCM, DPCM

Phrase

A term used in dictionary-based compression methods. It denotes a character sequence that gets encoded with a single code word. In the LZ78-family of compression methods, this term can be used as a synonim for a string.
Related link: Overview of dictionary methods

Pointer

A term used in the LZ77-family of compression methods. The encoding algorithm of the method encodes a re-occuring character sequence with a pointer which contains information about the location of the earlier occurence in the window.
Related term: lookahead buffer
Related link: LZ77: definition of terms

Predictive coding

A coding principle usually applied to the coding of signal values. The value of the next element is predicted based upon a certain combination of previous elements. The value stored represents the difference between the true and the predicted value.
Related terms: predictor, DPCM
Related link: DPCM: predictive coding

Predictor

A term used in predictive coding methods. It denotes the predicted value for the next data element.
Related term: DPCM
Related link: DPCM: predictive coding

Prefix

A term used in the LZ78-family of compression methods. It denotes a sequence of characters in the charstream which is followed by another character. This character is not a part of the prefix.
Related term: string
Related link: LZ78: definition of terms

Q

Quantization

The process of associating samples with their respective quantization segments. This results in the approximation of samples with quantization levels. This is a step in the process of signal digitizing.
Related terms: PCM, DPCM
Related link: PCM: introduction

Quantization, linear

The basic type of signal quantization with a constant quantization step size. The quantization noise depends only upon the quantization step size. The important measure for perceived signal quality is not the noise level, but the signal-to-noise ratio. With linear quantization, this ratio obviously drops for weaker signals, which results in poorer perceived quality.
Related term: PCM
Related link: PCM: example

Quantization, non-linear

A more advanced type of signal quantization. It uses variable quantization step size. This size gradually changes from fine at low amplitudes to coarse near maximum amplitudes. This results in less quantization noise for weaker signals, when compared to linear quantization with the same number of quantization levels. In principle, the levels are distributed in such a way that the signal-to-noise ratio remains constant across a wide dynamic range.
Related term: PCM
Related link: PCM: A-law compression

Quantization level

The value within a quantization segment with which all the sample values falling into it are approximated. This is the value the segment's code word is decoded to when the digital recording is being reproduced.
Related term: PCM
Related link: PCM: introduction

Quantization noise

The noise introduced to the digitized signal which results from the approximation of the samples with quantization levels. The level of this noise generally depends on the quantization step size.
Related terms: PCM, DPCM
Related link: PCM: signal restoration

Quantization segment

A range of sample values identified with a specific code word. A sample gets digitized when its value is associated with the code word of the segment into which it falls.
Related term: PCM
Related link: PCM: introduction

Quantization step size

The difference between the adjacent quantization levels. Smaller step size means less quantization noise. The step size can be uniform (linear quantization) or non-uniform (non-linear quantization). It can also be adaptive. Generally, more quantization levels means smaller step size.
Related terms: PCM, DPCM
Related link: PCM: reducing the number of quantization levels

S

Sample

The value of a signal recorded at one point in space or time. Taking samples of the signal at regular intervals is called signal sampling. It is used as the first step in signal digitizing.
Related terms: PCM, DPCM, quantization
Related link: PCM: introduction

Sampling frequency (fs)

The frequency at which the signal is sampled along the time or space axis. It determines the range of frequencies (bandwidth, B) in the input signal that can be recorded and reproduced: B = 0.5fs. For example, the audio signal can be sampled 44,100 times a second, which means the sampling frequency is 44.1 kHz, and the recording will contain all the frequencies of the input signal from 0 to 22.05 kHz.

Related link: PCM: reducing the number of samples

String

A term used in the LZ78-family of compression methods. It denotes a sequence of characters in the charstream, consisting of a prefix followed by a character.
Related link: LZ78: definition of terms

Substitutional methods

Another name for dictionary-based methods. The name comes from the basic encoding principle these methods use: they substitute phrases with code words.

U

u-law (mu-law) encoding/compression

A variant of the basic PCM encoding method, involving non-linear quantization. American equivalent of the A-law.

W

Window

A term used in the LZ77 compression method. It contains the previous data which has already been compressed. The window is searched for an earlier occurrence of the sequence that is just being compressed (contained in the lookahead buffer).
Related term: pointer
Related link: LZ77: definition of terms

 

 

HomePage - Basic Facts -Commercial -Algorithms - Hardware - FAQ - Related Links- Glossary

arcsepd.gif (196 bytes)

Maintained and Copyrighted 1997-1999 by Data Compression Reference Center

(compresswww@rasip.fer.hr)

arcsepd.gif (196 bytes)