"Source coding" redirects here. For the term in computer programming, seeSource code.
Ininformation theory,data compression,source coding,[1] orbit-rate reduction is the process of encodinginformation using fewerbits than the original representation.[2] Any particular compression is eitherlossy orlossless. Lossless compression reduces bits by identifying and eliminatingstatistical redundancy. No information is lost in lossless compression. Lossy compression reduces bits by removing unnecessary or less important information.[3] Typically, a device that performs data compression is referred to as an encoder, and one that performs the reversal of the process (decompression) as a decoder.
The process of reducing the size of adata file is often referred to as data compression. In the context ofdata transmission, it is called source coding: encoding is done at the source of the data before it is stored or transmitted.[4] Source coding should not be confused withchannel coding, for error detection and correction orline coding, the means for mapping data onto a signal.
Data Compression algorithms present aspace-time complexity trade-off between the bytes needed to store or transmit information, and theComputational resources needed to perform the encoding and decoding. The design of data compression schemes involves balancing the degree of compression, the amount of distortion introduced (when usinglossy data compression), and the computational resources or time required to compress and decompress the data.[5]
Lossless data compressionalgorithms usually exploitstatistical redundancy to represent data without losing anyinformation, so that the process is reversible. Lossless compression is possible because most real-world data exhibits statistical redundancy. For example, an image may have areas of color that do not change over several pixels; instead of coding "red pixel, red pixel, ..." the data may be encoded as "279 red pixels". This is a basic example ofrun-length encoding; there are many schemes to reduce file size by eliminating redundancy.
TheLempel–Ziv (LZ) compression methods are among the most popular algorithms for lossless storage.[6]DEFLATE is a variation on LZ optimized for decompression speed and compression ratio,[7] but compression can be slow. In the mid-1980s, following work byTerry Welch, theLempel–Ziv–Welch (LZW) algorithm rapidly became the method of choice for most general-purpose compression systems. LZW is used inGIF images, programs such asPKZIP, and hardware devices such as modems.[8] LZ methods use a table-based compression model where table entries are substituted for repeated strings of data. For most LZ methods, this table is generated dynamically from earlier data in the input. The table itself is oftenHuffman encoded.Grammar-based codes like this can compress highly repetitive input extremely effectively, for instance, a biologicaldata collection of the same or closely related species, a huge versioned document collection, internet archival, etc. The basic task of grammar-based codes is constructing a context-free grammar deriving a single string. Other practical grammar compression algorithms includeSequitur andRe-Pair.
The strongest modern lossless compressors useprobabilistic models, such asprediction by partial matching. TheBurrows–Wheeler transform can also be viewed as an indirect form of statistical modelling.[citation needed] In a further refinement of the direct use ofprobabilistic modelling, statistical estimates can be coupled to an algorithm calledarithmetic coding. Arithmetic coding is a more modern coding technique that uses the mathematical calculations of afinite-state machine to produce a string of encoded bits from a series of input data symbols. It can achieve superior compression compared to other techniques such as the better-known Huffman algorithm. It uses an internal memory state to avoid the need to perform a one-to-one mapping of individual input symbols to distinct representations that use an integer number of bits, and it clears out the internal memory only after encoding the entire string of data symbols. Arithmetic coding applies especially well to adaptive data compression tasks where the statistics vary and are context-dependent, as it can be easily coupled with an adaptive model of theprobability distribution of the input data. An early example of the use of arithmetic coding was in an optional (but not widely used) feature of theJPEG image coding standard.[9] It has since been applied in various other designs includingH.263,H.264/MPEG-4 AVC andHEVC for video coding.[10]
Archive software typically has the ability to adjust the "dictionary size", where a larger size demands morerandom-access memory during compression and decompression, but compresses stronger, especially on repeating patterns in files' content.[11][12]
Composite image showing JPG and PNG image compression. Left side of the image is from a JPEG image, showing lossy artifacts; the right side is from a PNG image.
In the late 1980s, digital images became more common, and standards for losslessimage compression emerged. In the early 1990s, lossy compression methods began to be widely used.[13] In these schemes, some loss of information is accepted as dropping nonessential detail can save storage space. There is a correspondingtrade-off between preserving information and reducing size. Lossy data compression schemes are designed by research on how people perceive the data in question. For example, the human eye is more sensitive to subtle variations inluminance than it is to the variations in color. JPEG image compression works in part by rounding off nonessential bits of information.[14] A number of popular compression formats exploit these perceptual differences, includingpsychoacoustics for sound, andpsychovisuals for images and video.
Most forms of lossy compression are based ontransform coding, especially thediscrete cosine transform (DCT). It was first proposed in 1972 byNasir Ahmed, who then developed a working algorithm with T. Natarajan andK. R. Rao in 1973, before introducing it in January 1974.[15][16] DCT is the most widely used lossy compression method, and is used in multimedia formats for images (such as JPEG andHEIF),[17]video (such asMPEG,AVC and HEVC) and audio (such asMP3,AAC andVorbis).
In lossy audio compression, methods of psychoacoustics are used to remove non-audible (or less audible) components of theaudio signal. Compression of human speech is often performed with even more specialized techniques;speech coding is distinguished as a separate discipline from general-purpose audio compression. Speech coding is used ininternet telephony, for example, audio compression is used for CD ripping and is decoded by the audio players.[citation needed]
There is a close connection betweenmachine learning and compression. A system that predicts theposterior probabilities of a sequence given its entire history can be used for optimal data compression (by usingarithmetic coding on the output distribution). Conversely, an optimal compressor can be used for prediction (by finding the symbol that compresses best, given the previous history). This equivalence has been used as a justification for using data compression as a benchmark for "general intelligence".[19][20][21]
An alternative view can show compression algorithms implicitly map strings into implicitfeature space vectors, and compression-based similarity measures compute similarity within these feature spaces. For each compressor C(.) we define an associated vector space ℵ, such that C(.) maps an input string x, corresponding to the vector norm ||~x||. An exhaustive examination of the feature spaces underlying all compression algorithms is precluded by space; instead, feature vectors chooses to examine three representative lossless compression methods, LZW, LZ77, and PPM.[22]
According toAIXI theory, a connection more directly explained inHutter Prize, the best possible compression of x is the smallest possible software that generates x. For example, in that model, a zip file's compressed size includes both the zip file and the unzipping software, since you can not unzip it without both, but there may be an even smaller combined form.
Examples of AI-powered audio/video compression software includeNVIDIA Maxine, AIVC.[23] Examples of software that can perform AI-powered image compression includeOpenCV,TensorFlow,MATLAB's Image Processing Toolbox (IPT) and High-Fidelity Generative Image Compression.[24]
Inunsupervised machine learning,k-means clustering can be utilized to compress data by grouping similar data points into clusters. This technique simplifies handling extensive datasets that lack predefined labels and finds widespread use in fields such asimage compression.[25]
Data compression aims to reduce the size of data files, enhancing storage efficiency and speeding up data transmission. K-means clustering, an unsupervised machine learning algorithm, is employed to partition a dataset into a specified number of clusters, k, each represented by thecentroid of its points. This process condenses extensive datasets into a more compact set of representative points. Particularly beneficial inimage andsignal processing, k-means clustering aids in data reduction by replacing groups of data points with their centroids, thereby preserving the core information of the original data while significantly decreasing the required storage space.[26]
Large language models (LLMs) are also efficient lossless data compressors on some data sets, as demonstrated byDeepMind's research with the Chinchilla 70B model. Developed by DeepMind, Chinchilla 70B effectively compressed data, outperforming conventional methods such asPortable Network Graphics (PNG) for images andFree Lossless Audio Codec (FLAC) for audio. It achieved compression of image and audio data to 43.4% and 16.4% of their original sizes, respectively. There is, however, some reason to be concerned that the data set used for testing overlaps the LLM training data set, making it possible that the Chinchilla 70B model is only an efficient compression tool on data it has already been trained on.[27][28]
Data compression can be viewed as a special case ofdata differencing.[29][30] Data differencing consists of producing adifference given asource and atarget, with patching reproducing thetarget given asource and adifference. Since there is no separate source and target in data compression, one can consider data compression as data differencing with empty source data, the compressed file corresponding to a difference from nothing. This is the same as considering absoluteentropy (corresponding to data compression) as a special case ofrelative entropy (corresponding to data differencing) with no initial data.
The termdifferential compression is used to emphasize the data differencing connection.
An important image compression technique is thediscrete cosine transform (DCT), a technique developed in the early 1970s.[15] DCT is the basis for JPEG, alossy compression format which was introduced by theJoint Photographic Experts Group (JPEG) in 1992.[34] JPEG greatly reduces the amount of data required to represent an image at the cost of a relatively small reduction in image quality and has become the most widely usedimage file format.[35][36] Its highly efficient DCT-based compression algorithm was largely responsible for the wide proliferation ofdigital images anddigital photos.[37]
Lossy audio compression algorithms provide higher compression and are used in numerous audio applications includingVorbis andMP3. These algorithms almost all rely onpsychoacoustics to eliminate or reduce fidelity of less audible sounds, thereby reducing the space required to store or transmit them.[2][46]
The acceptable trade-off between loss of audio quality and transmission or storage size depends upon the application. For example, one 640 MBcompact disc (CD) holds approximately one hour of uncompressedhigh fidelity music, less than 2 hours of music compressed losslessly, or 7 hours of music compressed in theMP3 format at a mediumbit rate. A digital sound recorder can typically store around 200 hours of clearly intelligible speech in 640 MB.[47]
Lossless audio compression produces a representation of digital data that can be decoded to an exact digital duplicate of the original. Compression ratios are around 50–60% of the original size,[48] which is similar to those for generic lossless data compression. Lossless codecs usecurve fitting or linear prediction as a basis for estimating the signal. Parameters describing the estimation and the difference between the estimation and the actual signal are coded separately.[49]
Someaudio file formats feature a combination of a lossy format and a lossless correction; this allows stripping the correction to easily obtain a lossy file. Such formats includeMPEG-4 SLS (Scalable to Lossless),WavPack, andOptimFROG DualStream.
When audio files are to be processed, either by further compression or forediting, it is desirable to work from an unchanged original (uncompressed or losslessly compressed). Processing of a lossily compressed file for some purpose usually produces a final result inferior to the creation of the same compressed file from an uncompressed original. In addition to sound editing or mixing, lossless audio compression is often used for archival storage, or as master copies.
Comparison ofspectrograms of audio in an uncompressed format and several lossy formats. The lossy spectrograms showbandlimiting of higher frequencies, a common technique associated with lossy audio compression.
Lossy audio compression is used in a wide range of applications. In addition to standalone audio-only applications of file playback in MP3 players or computers, digitally compressed audio streams are used in most video DVDs, digital television, streaming media on theInternet, satellite and cable radio, and increasingly in terrestrial radio broadcasts. Lossy compression typically achieves far greater compression than lossless compression, by discarding less-critical data based onpsychoacoustic optimizations.[50]
Psychoacoustics recognizes that not all data in an audio stream can be perceived by the humanauditory system. Most lossy compression reduces redundancy by first identifying perceptually irrelevant sounds, that is, sounds that are very hard to hear. Typical examples include high frequencies or sounds that occur at the same time as louder sounds. Those irrelevant sounds are coded with decreased accuracy or not at all.
Due to the nature of lossy algorithms,audio quality suffers adigital generation loss when a file is decompressed and recompressed. This makes lossy compression unsuitable for storing the intermediate results in professional audio engineering applications, such as sound editing and multitrack recording. However, lossy formats such asMP3 are very popular with end-users as the file size is reduced to 5-20% of the original size and a megabyte can store about a minute's worth of music at adequate quality.
Several proprietary lossy compression algorithms have been developed that provide higher quality audio performance by using a combination of lossless and lossy algorithms with adaptive bit rates and lower compression ratios. Examples includeaptX,LDAC,LHDC,MQA andSCL6.
To determine what information in an audio signal is perceptually irrelevant, most lossy compression algorithms use transforms such as themodified discrete cosine transform (MDCT) to converttime domain sampled waveforms into a transform domain, typically thefrequency domain. Once transformed, component frequencies can be prioritized according to how audible they are. Audibility of spectral components is assessed using theabsolute threshold of hearing and the principles ofsimultaneous masking—the phenomenon wherein a signal is masked by another signal separated by frequency—and, in some cases,temporal masking—where a signal is masked by another signal separated by time.Equal-loudness contours may also be used to weigh the perceptual importance of components. Models of the human ear-brain combination incorporating such effects are often calledpsychoacoustic models.[51]
Other types of lossy compressors, such as thelinear predictive coding (LPC) used with speech, are source-based coders. LPC uses a model of the human vocal tract to analyze speech sounds and infer the parameters used by the model to produce them moment to moment. These changing parameters are transmitted or stored and used to drive another model in the decoder which reproduces the sound.
Lossy formats are often used for the distribution of streaming audio or interactive communication (such as in cell phone networks). In such applications, the data must be decompressed as the data flows, rather than after the entire data stream has been transmitted. Not all audio codecs can be used for streaming applications.[50]
Latency is introduced by the methods used to encode and decode the data. Some codecs will analyze a longer segment, called aframe, of the data to optimize efficiency, and then code it in a manner that requires a larger segment of data at one time to decode. The inherent latency of the coding algorithm can be critical; for example, when there is a two-way transmission of data, such as with a telephone conversation, significant delays may seriously degrade the perceived quality.
In contrast to the speed of compression, which is proportional to the number of operations required by the algorithm, here latency refers to the number of samples that must be analyzed before a block of audio is processed. In the minimum case, latency is zero samples (e.g., if the coder/decoder simply reduces the number of bits used to quantize the signal). Time domain algorithms such as LPC also often have low latencies, hence their popularity in speech coding for telephony. In algorithms such as MP3, however, a large number of samples have to be analyzed to implement a psychoacoustic model in the frequency domain, and latency is on the order of 23 ms.
Speech encoding is an important category of audio data compression. The perceptual models used to estimate what aspects of speech a human ear can hear are generally somewhat different from those used for music. The range of frequencies needed to convey the sounds of a human voice is normally far narrower than that needed for music, and the sound is normally less complex. As a result, speech can be encoded at high quality using a relatively low bit rate.
This is accomplished, in general, by some combination of two approaches:
Only encoding sounds that could be made by a single human voice.
Throwing away more of the data in the signal—keeping just enough to reconstruct an "intelligible" voice rather than the full frequency range of humanhearing.
The earliest algorithms used in speech encoding (and audio data compression in general) were theA-law algorithm and theμ-law algorithm.
The world's first commercialbroadcast automation audio compression system was developed by Oscar Bonello, an engineering professor at theUniversity of Buenos Aires.[63]In 1983, using the psychoacoustic principle of the masking of critical bands first published in 1967,[64] he started developing a practical application based on the recently developedIBM PC computer, and the broadcast automation system was launched in 1987 under the nameAudicom.[65]35 years later, almost all the radio stations in the world were using this technology manufactured by a number of companies because the inventor refused to patent his work, preferring to publish it and leave it in the public domain.[66]
A literature compendium for a large variety of audio coding systems was published in the IEEE'sJournal on Selected Areas in Communications (JSAC), in February 1988. While there were some papers from before that time, this collection documented an entire variety of finished, working audio coders, nearly all of them using perceptual techniques and some kind of frequency analysis and back-end noiseless coding.[67]
The two key video compression techniques used invideo coding standards are the DCT andmotion compensation (MC). Most video coding standards, such as theH.26x andMPEG formats, typically use motion-compensated DCT video coding (block motion compensation).[69][70]
Most video codecs are used alongside audio compression techniques to store the separate but complementary data streams as one combined package using so-calledcontainer formats.[71]
Video data may be represented as a series of still image frames. Such data usually contains abundant amounts of spatial and temporalredundancy. Video compression algorithms attempt to reduce redundancy and store information more compactly.
Mostvideo compression formats andcodecs exploit both spatial and temporal redundancy (e.g. through difference coding withmotion compensation). Similarities can be encoded by only storing differences between e.g. temporally adjacent frames (inter-frame coding) or spatially adjacent pixels (intra-frame coding).Inter-frame compression (a temporaldelta encoding) (re)uses data from one or more earlier or later frames in a sequence to describe the current frame.Intra-frame coding, on the other hand, uses only data from within the current frame, effectively being still-image compression.[51]
Theintra-frame video coding formats used in camcorders and video editing employ simpler compression that uses only intra-frame prediction. This simplifies video editing software, as it prevents a situation in which a compressed frame refers to data that the editor has deleted.
Usually, video compression additionally employslossy compression techniques likequantization that reduce aspects of the source data that are (more or less) irrelevant to the human visual perception by exploiting perceptual features of human vision. For example, small differences in color are more difficult to perceive than are changes in brightness. Compression algorithms can average a color across these similar areas in a manner similar to those used in JPEG image compression.[9] As in all lossy compression, there is atrade-off betweenvideo quality andbit rate, cost of processing the compression and decompression, and system requirements. Highly compressed video may present visible or distractingartifacts.
Other methods other than the prevalent DCT-based transform formats, such asfractal compression,matching pursuit and the use of adiscrete wavelet transform (DWT), have been the subject of some research, but are typically not used in practical products.Wavelet compression is used in still-image coders and video coders without motion compensation. Interest in fractal compression seems to be waning, due to recent theoretical analysis showing a comparative lack of effectiveness of such methods.[51]
In inter-frame coding, individual frames of a video sequence are compared from one frame to the next, and thevideo compression codec records thedifferences to the reference frame. If the frame contains areas where nothing has moved, the system can simply issue a short command that copies that part of the previous frame into the next one. If sections of the frame move in a simple manner, the compressor can emit a (slightly longer) command that tells the decompressor to shift, rotate, lighten, or darken the copy. This longer command still remains much shorter than data generated by intra-frame compression. Usually, the encoder will also transmit a residue signal which describes the remaining more subtle differences to the reference imagery. Using entropy coding, these residue signals have a more compact representation than the full signal. In areas of video with more motion, the compression must encode more data to keep up with the larger number of pixels that are changing. Commonly during explosions, flames, flocks of animals, and in some panning shots, the high-frequency detail leads to quality decreases or to increases in thevariable bitrate.
Many commonly used video compression methods (e.g., those in standards approved by theITU-T orISO) share the same basic architecture that dates back toH.261 which was standardized in 1988 by the ITU-T. They mostly rely on the DCT, applied to rectangular blocks of neighboring pixels, and temporal prediction usingmotion vectors, as well as nowadays also an in-loop filtering step.
In the prediction stage, variousdeduplication and difference-coding techniques are applied that help decorrelate data and describe new data based on already transmitted data.
Then rectangular blocks of remainingpixel data are transformed to the frequency domain. In the main lossy processing stage, frequency domain data gets quantized in order to reduce information that is irrelevant to human visual perception.
In the last stage statistical redundancy gets largely eliminated by anentropy coder which often applies some form of arithmetic coding.
In an additional in-loop filtering stage various filters can be applied to the reconstructed image signal. By computing these filters also inside the encoding loop they can help compression because they can be applied to reference material before it gets used in the prediction process and they can be guided using the original signal. The most popular example aredeblocking filters that blur out blocking artifacts from quantization discontinuities at transform block boundaries.
In 1967, A.H. Robinson and C. Cherry proposed arun-length encoding bandwidth compression scheme for the transmission of analog television signals.[72] The DCT, which is fundamental to modern video compression,[73] was introduced byNasir Ahmed, T. Natarajan andK. R. Rao in 1974.[16][74]
H.261, which debuted in 1988, commercially introduced the prevalent basic architecture of video compression technology.[75] It was the firstvideo coding format based on DCT compression.[73] H.261 was developed by a number of companies, includingHitachi,PictureTel,NTT,BT andToshiba.[76]
Genetics compression algorithms are the latest generation of lossless algorithms that compress data (typically sequences of nucleotides) using both conventional compression algorithms and genetic algorithms adapted to the specific datatype. In 2012, a team of scientists from Johns Hopkins University published a genetic compression algorithm that does not use a reference genome for compression. HAPZIPPER was tailored forHapMap data and achieves over 20-fold compression (95% reduction in file size), providing 2- to 4-fold better compression and is less computationally intensive than the leading general-purpose compression utilities. For this, Chanda, Elhaik, and Bader introduced MAF-based encoding (MAFE), which reduces the heterogeneity of the dataset by sorting SNPs by their minor allele frequency, thus homogenizing the dataset.[80] Other algorithms developed in 2009 and 2013 (DNAZip and GenomeZip) have compression ratios of up to 1200-fold—allowing 6 billion basepair diploid human genomes to be stored in 2.5 megabytes (relative to a reference genome or averaged over many genomes).[81][82] For a benchmark in genetics/genomics data compressors, see[83]
It is estimated that the total amount of data that is stored on the world's storage devices could be further compressed with existing compression algorithms by a remaining average factor of 4.5:1.[84] It is estimated that the combined technological capacity of the world to store information provides 1,300exabytes of hardware digits in 2007, but when the corresponding content is optimally compressed, this only represents 295 exabytes ofShannon information.[85]
^Wade, Graham (1994).Signal coding and processing (2 ed.). Cambridge University Press. p. 34.ISBN978-0-521-42336-6. Retrieved2011-12-22.The broad objective of source coding is to exploit or remove 'inefficient' redundancy in thePCM source and thereby achieve a reduction in the overall source rate R.
^Salomon, David (2008).A Concise Introduction to Data Compression. Berlin: Springer.ISBN9781848000728.
^Tank, M.K. (2011). "Implementation of Lempel-ZIV algorithm for lossless compression using VHDL".Thinkquest 2010: Proceedings of the First International Conference on Contours of Computing Technology. Berlin: Springer. pp. 275–283.doi:10.1007/978-81-8489-989-4_51.ISBN978-81-8489-988-7.
^Korn, D.G.; Vo, K.P. (1995). B. Krishnamurthy (ed.).Vdelta: Differencing and Compression. Practical Reusable Unix Software. New York: John Wiley & Sons, Inc.
^Hoffman, Roy (2012).Data Compression in Digital Systems.Springer Science & Business Media. p. 124.ISBN9781461560319.Basically, wavelet coding is a variant on DCT-based transform coding that reduces or eliminates some of its limitations. (...) Another advantage is that rather than working with 8 × 8 blocks of pixels, as do JPEG and other block-based DCT techniques, wavelet coding can simultaneously compress the entire image.
^The Olympus WS-120 digital speech recorder, according to its manual, can store about 178 hours of speech-quality audio in .WMA format in 500 MB of flash memory.
^US patent 2605361, C. Chapin Cutler, "Differential Quantization of Communication Signals", issued 1952-07-29
^Cummiskey, P.; Jayant, N. S.; Flanagan, J. L. (1973). "Adaptive Quantization in Differential PCM Coding of Speech".Bell System Technical Journal.52 (7):1105–1118.doi:10.1002/j.1538-7305.1973.tb02007.x.
^Cummiskey, P.; Jayant, Nikil S.; Flanagan, J. L. (1973). "Adaptive quantization in differential PCM coding of speech".The Bell System Technical Journal.52 (7):1105–1118.doi:10.1002/j.1538-7305.1973.tb02007.x.ISSN0005-8580.
^abcSchroeder, Manfred R. (2014)."Bell Laboratories".Acoustics, Information, and Communication: Memorial Volume in Honor of Manfred R. Schroeder. Springer. p. 388.ISBN9783319056609.
^Britanak, V. (2011). "On Properties, Relations, and Simplified Implementation of Filter Banks in the Dolby Digital (Plus) AC-3 Audio Coding Standards".IEEE Transactions on Audio, Speech, and Language Processing.19 (5):1231–1241.doi:10.1109/TASL.2010.2087755.S2CID897622.
^Princen, J.; Johnson, A.; Bradley, A. (1987). "Subband/Transform coding using filter bank designs based on time domain aliasing cancellation".ICASSP '87. IEEE International Conference on Acoustics, Speech, and Signal Processing. Vol. 12. pp. 2161–2164.doi:10.1109/ICASSP.1987.1169405.S2CID58446992.
^Princen, J.; Bradley, A. (1986). "Analysis/Synthesis filter bank design based on time domain aliasing cancellation".IEEE Transactions on Acoustics, Speech, and Signal Processing.34 (5):1153–1161.doi:10.1109/TASSP.1986.1164954.
^Dmitriy Vatolin; et al. (Graphics & Media Lab Video Group) (March 2007).Lossless Video Codecs Comparison '2007(PDF) (Report). Moscow State University.Archived(PDF) from the original on 2008-05-15.
^"Video Coding".CSIP website. Center for Signal and Information Processing, Georgia Institute of Technology. Archived fromthe original on 23 May 2013. Retrieved6 March 2013.
