Editing Fax (section)

=== Compression ===
As well as specifying the resolution (and allowable physical size) of the image being faxed, the ITU-T&nbsp;T.4 recommendation specifies two compression methods for decreasing the amount of data that needs to be transmitted between the fax machines to transfer the image. The two methods defined in T.4 are:<ref name="T4">{{cite web |url=http://www.itu.int/rec/T-REC-T.4/en |title=T.4: Standardization of Group 3 facsimile terminals for document transmission |publisher=ITU-T |date=2011-03-14 |access-date=2013-12-28}}</ref>
* [[Modified Huffman coding|Modified Huffman]] (MH).
* [[Modified READ]] (MR) (''Relative Element Address Designate''<ref>{{cite journal | url=https://ieeexplore.ieee.org/document/1456020 | doi=10.1109/PROC.1980.11751 | title=International digital facsimile coding standards | year=1980 | last1=Hunter | first1=R. | last2=Robinson | first2=A.H. | journal=Proceedings of the IEEE | volume=68 | issue=7 | pages=854–867 | s2cid=46403372 }}</ref>), optional.
An additional method is specified in T.6:<ref name="T6">{{cite web| url=http://www.itu.int/rec/T-REC-T.6/en |title=T.6: Facsimile coding schemes and coding control functions for Group 4 facsimile apparatus |publisher=ITU-T |date=November 1988 |access-date=2013-12-28}}</ref>
* [[Group 4 compression|Modified Modified READ]] (MMR).
Later, other compression techniques were added as options to ITU-T recommendation T.30, such as the more efficient [[JBIG]] (T.82, T.85) for bi-level content, and [[JPEG]] (T.81), T.43, [[Mixed raster content|MRC]] (T.44), and T.45 for grayscale, palette, and colour content.<ref name="T30">{{cite web |url=http://www.itu.int/rec/T-REC-T.30 |title=T.30: Procedures for document facsimile transmission in the general switched telephone network |publisher=ITU-T |date=2014-05-15 |access-date=2013-12-28}}</ref> Fax machines can negotiate at the start of the T.30 session to use the best technique implemented on both sides.

==== Modified Huffman ====
{{Main|Modified Huffman coding}}
Modified Huffman (MH), specified in T.4 as the one-dimensional coding scheme, is a codebook-based run-length encoding scheme optimised to efficiently compress whitespace.<ref name="T4"/> As most faxes consist mostly of white space, this minimises the transmission time of most faxes. Each line scanned is compressed independently of its predecessor and successor.<ref name="T4"/>

==== Modified READ ====
Modified READ, specified as an optional two-dimensional coding scheme in T.4, encodes the first scanned line using MH.<ref name="T4"/> The next line is compared to the first, the differences determined, and then the differences are encoded and transmitted.<ref name="T4"/> This is effective, as most lines differ little from their predecessor.  This is not continued to the end of the fax transmission, but only for a limited number of lines until the process is reset, and a new "first line" encoded with MH is produced. This limited number of lines is to prevent errors propagating throughout the whole fax, as the standard does not provide for error correction. This is an optional facility, and some fax machines do not use MR in order to minimise the amount of computation required by the machine. The limited number of lines is 2 for "Standard"-resolution faxes, and 4 for "Fine"-resolution faxes.

==== Modified Modified READ ====
{{Main|Group 4 compression}}
The ITU-T T.6 recommendation adds a further compression type of Modified Modified READ (MMR), which simply allows a greater number of lines to be coded by MR than in T.4.<ref name="T6"/> This is because T.6 makes the assumption that the transmission is over a circuit with a low number of line errors, such as digital [[Integrated Services Digital Network|ISDN]]. In this case, the number of lines for which the differences are encoded is not limited.

==== JBIG ====

In 1999, ITU-T recommendation T.30 added [[JBIG]] (ITU-T T.82) as another [[lossless]] [[Binary image|bi-level]] compression algorithm, or more precisely a "fax profile" subset of JBIG (ITU-T T.85). JBIG-compressed pages result in 20% to 50% faster transmission than MMR-compressed pages, and up to 30 times faster transmission if the page includes [[halftone]] images.

[[JBIG]] performs [[adaptive compression]], that is, both the encoder and decoder collect statistical information about the transmitted image from the pixels transmitted so far, in order to predict the probability for each next pixel being either black or white. For each new pixel, JBIG looks at ten nearby, previously transmitted pixels. It counts, how often in the past the next pixel has been black or white in the same neighborhood, and estimates from that the probability distribution of the next pixel. This is fed into an [[arithmetic coder]], which adds only a small fraction of a bit to the output sequence if the more probable pixel is then encountered.

The ITU-T T.85 "fax profile" constrains some optional features of the full JBIG standard, such that codecs do not have to keep data about more than the last three pixel rows of an image in memory at any time. This allows the streaming of "endless" images, where the height of the image may not be known until the last row is transmitted.

ITU-T T.30 allows fax machines to negotiate one of two options of the T.85 "fax profile":
* In "basic mode", the JBIG encoder must split the image into horizontal stripes of 128 lines (parameter L0&nbsp;=&nbsp;128) and restart the arithmetic encoder for each stripe.
* In "option mode", there is no such constraint.

==== Matsushita Whiteline Skip ====
A proprietary compression scheme employed on Panasonic fax machines is Matsushita Whiteline Skip (MWS). It can be overlaid on the other compression schemes, but is operative only when two Panasonic machines are communicating with one another. This system detects the blank scanned areas between lines of text, and then compresses several blank scan lines into the data space of a single character. (JBIG implements a similar technique called "typical prediction", if header flag TPBON is set to 1.)