1-D signal - A signal that its value depends only on one independent variables. Such a signal is an audio signal (its value depends on the time).

2-D signal - A signal that its value depends on two independent variables. Such a signal is an image (its value depends on the width and the height).

Analog signal - A signal that takes analog values e.g. electric voltage or electric current intensity.

AR modelling - A power spectrum estimation algorithm driven by a white noise process.

Autocorrelation - A function that estimates the product of a signal by its complex conjugate.

Blackman-Tukey estimator - A power spectrum estimator that produces smooth estimation by using a suitable window.

Circular convolution - A convolution were the delay variables are calculated circularly, by using modΝ, where Ν is the length of the signal.

Cosal prediction - The prediction of a pixel luminance value based on the pixels luminance values that have been already calculated.

Digitization - The process that converts a signal (audio, image etc.) from analog to digital.

Distinct signal - A signal that takes values only for distinct values of its variables. It is usually a result of an analog signal sampling.

Edge detector mask - A matrix that contains weight vectors that are used to detect edges towards certain directions.

Edge detection operators - Arithmetic operators that when they are applied on an image or a part of an image they detect its edges.

Edge following - The process that is applied on an edge image in order to detect connected edges.

Edge thresholding - The process that follows an edge detector algorithm and decides whether an edge is present or not. The thresholding can be based on local or global information.

Entropy coding - A coding algorithm based on the coding of the luminance of an image.

FFT Butterfly - The elementary calculation in FFT.

Filter - A 2-D discrete system that transforms a signal (input) to another signal (output).

Histogram - A function that shows the luminance values distribution of an image.

Hough Transform-An algorithm that detects small geometrical shapes in edge images.

Huffman coding - A coding algorithm that uses codewords to replace the most probable bit sequences with codewords of small length.

Huffman tree - A tree structure that is used to construct the Huffman code.

Image edge -The border between two homogeneous regions of an image, which have different luminance values.

Impulse response - The output of a system when its input is an impulse.

Impulsive noise - Noise that appears as white or black pixels.

Image entropy - A measure of the information contained in an image.

Image compression - Techniques that are used to reduce the memory needed for the storage of digital images.

Lowpass signal - A signal that its spectrum has a lot of energy around the origin (low frequencies).

LZW algorithm - A lossless compression algorithm based on the formation of a matrix that contains codewords. Each codeword corresponds to a bit sequence. The more probable a bit sequence is, the shorter the codeword should be.

Laplace operator - It is defined based on the second order partial dirivants of an image function. It is used as an edge detector.

Luminance local variance - A measure for the homogeneity of an image region. When the local variance is small the region is considered homogeneous.

Modified READ coding - Coding that takes into account vertical correlations among run transitions in consecutive image lines.

Modified Huffman code - Modification of the Huffman code based on statistics observed when typical documents were transmitted using FAX.

Multidimensional signal - The signal that its value depends on more than two independent variables.

Neighboring pixel - A pixel that is spatially next to another pixel.

Prediction coding
- A coding algorithm that takes into account the spatial correlation of pixel values in order to predict neighboring pixel values.

Predictive differential pulse code modulation - Coding using prediction coefficients and coded prediction error.

Optical detector - The mechanism for digital image formation. There are two types; the photochemical (film) and the photoelectric (camera).

Photoelectric detector - It is a digital image formation mechanism that is used in cameras. Types of photoelectric detectors are Vidicon, Charge Injection Denices (CID), Charge Coupled Devices (CCD) cameras.

PCM modulation - A coding scheme that uses a uniform quantizer.

Prediction window - A small part of a signal (neighbor) that is used for prediction.

Periodogram - The Fourier transform of the autocorrelation function.

Predictability - The ability to predict the luminance of a pixel based on the luminance values of its neighborhood.

Prediction coefficients - The coefficients of the prediction window that can be selected to optimise the prediction of the signal.

Quantization - The process that substitutes the values of a signal with approximations which are contained in a set of values.

Row column FFT - An algorithm that calculates DFT by using 1D DFTs over rows and columns.

Run length coding - A coding algorithm that uses the number of sequential pixels with equal luminance.

Spectrum distortion - Any change in the spectrum of a signal (e.g. if certain frequencies are removed)

Sample - The value of a signal at a certain moment (e.g. for an audio signal) or at a certain point (e.g. for an image).

Sampling interval - The time interval between succesive samples during the sampling process.

Separable signal - A 2-D signal that can be expressed as a product of two 1-D signals.

Sampling frequency - The frequency that is used to sample an analog signal and produce a digital signal.

Spectrum - The Fourier transformation of an image or a signal in a frequency domain.

Support region - The orthogonal region where a discrete signal takes non zero values.

Transform image coding - Any coding that uses coefficients of an image transformation, in order to compress an image.

Transition pixel - A pixel in the reference line or the coding line that indicates a luminance change (used in modified READ coding).

Uniform sampling - The sampling process with equal sampling intervals (e.g. one sample every one second).

Vector Radix FFT - An algorithm that calculates DFT by using smaller 2-D DFTs.

White noise - The noise that spreads equally in all frequencies.