FAX IMAGE CODING SCHEMES (VoIP)

14.2
An optical scanner scans the document and generates a series of electrical signals corresponding to the picture elements on the scan line. On a standard size (8.5″ x 11″) document, the scanner gives 1728 bits of data or pixels in the horizontal direction and 1145 lines of information in the vertical direction (while using the lowest vertical resolution that is 3.85 lines per mm). This procedure produces (1728) x (1145) = 1,978,560 bits (approximately 2 million bits per page) of information from one page. Without any compression and coding techniques, it would take minimum of 207 seconds to transmit the entire one page data at 9600 bps rate [URL (Maine)].
The method that G3 fax machines use to reduce data is called coding. In voice, voice compression or codecs are used for the compression operation. In fax, the coding name is used for compression. Fax coding performs data compression, and voice codecs perform signal compression. The ITU recommendations T.4 and T.6 are used for coding the fax data. T.4 implementation comprises image information, encoding, and decoding operations that are required for a facsimile device. It also considers image size and resolution. T.4 and T.30 complement each other in a G3 facsimile device. The main parts of T.4 are the run-length coding/decoding schemes, the bilevel compression schemes, and the color/grayscale modes. Bilevel scan lines are composed of white and black areas, and groups of black and white pixels are placed in every line of picture data. These pixel groups are referred to as black run lengths and white run lengths of variable size. Based on the run-length size, identifying code words are assigned to the different size run lengths and only these codes are sent to save the bit rate for transmission. The popular coding scheme used with G3 fax machines are the 1- D MH scheme and the 2- D MR scheme
[McConnell et al. (1999), URL (Fax-theory)]. MH and MR are used in many G3 fax machines. The most recent G3 fax machines use an optional MMR scheme [ITU-T-T.6 (1988)]. A G3 color fax machine uses joint photographic expert group (JPEG) and joint bilevel image experts group (JBIG) color-coding schemes for coding the color pages. Group-4 devices follow the T.6 recommendation for image compression and coding. G4 uses MMR, trellis, and JBIG coding schemes. These coding schemes of MMR, JPEG, and JBIG color-coding are incorporated into recent versions of G3 fax machines, mainly in SG3 facsimile devices. Each one of these coding methods reduces the amount of data needed to be sent over the phone lines, and each improvement in coding yields an improvement in data transmission speed. Typical compression achieved with fax coding techniques is given here [URL (Multitech)- : Modified Huffman 5:1, Modified read 7.5:1, and Modified modified read 10:1, JBIG and JPEG 10:1 to 20:1.
14.2.1


Modified Huffman 1-D Coding

Modified Huffman is a 1-D coding scheme, which is supported as mandatory in all G3 facsimile devices. An overview on MH is given in this section. MH is the most common black-and-white encoding scheme, in which one line of data is scanned and coded for transmission. An optional ECM can also be used to transmit the total coded scan line in an MH coding scheme. This encoding scheme compresses a typical facsimile black-and-white image six to ten times. MH coding uses two different look-up tables of codes called terminating codes and make-up codes. An MH coded line always starts with a white run length. If an actual scan line begins with a black run length, a white run length of zero is sent. The run length can be either a termination or a make-up code word. If it is a terminating code word, then the next run length will be black. If it is a make-up code, then the next run length will be white. Each coded line is terminated with an end-of-line (EOL) code. The synchronized EOL code word consists of 11 zeros followed by one bit of "1." The end of the facsimile document page is marked with six consecutive EOL codes. The pattern of six consecutive EOLs is called return to control (RTC). The longer the run lengths of black and white, the compression becomes better. Each picture element is represented by either one terminating code word or one make-up code word followed by a terminating code word. Run lengths from 0 to 63 are coded using the appropriate terminating code words. Run lengths in the range 64 to 1728 are first coded using the appropriate make-up code word followed by the terminating code word, which represents the difference between the run length and the make-up code word. Different code words are available for both black and white run lengths. A list of code words for black and white run lengths is given in T.4.
EOL: Each coded line of data is followed by an EOL. The EOL format is
of 12 bits with the binary value 000000000001 (11 zeros and a one). The
beginning of page transmission is indicated by EOL. This unique code word is not found in a valid line of data. Hence, resynchronization is possible even in the presence of error burst.
Fill: The fill bits are used for creating pause (silence) in the message flow. Fill bits are of a variable-length string of zeros. Fill can be inserted between a line of data and an EOL but never within a line of data. A fill is inserted to ensure that the transmission time of data, EOL, and fill is greater than the minimum transmission time of the total coded scan line but less than 5 seconds [ITU-T-T.4 (2003)].
RTC: The format for RTC is EOL repeated for six times (EOL EOL EOL EOL EOL EOL). These six EOLs indicate end-of-document transmission. After the return to control signal, the transmitter will send postmes-sage commands in a framed format as well as the data-signaling rate of
the controlled signals [ITU - T- T.4 (2003) ].
14.2.2

Modified Read (MR) 2-D Scheme

The MR coding scheme is also known as two-dimensional coding that works very similar to the modified Huffman scheme. This coding/decoding algorithm is explained in the T.4 recommendation [ITU-T-T.4 (2003)] and is popularly used in G3 machines. MR gives higher compression than MH. Similar to the MH coding scheme, MR uses an optional ECM for transmitting the coded scan line. The synchronization code consists of EOL (11 zeros followed by a bit of “1″) plus a tag bit. The tag bit indicates whether the following line is coded as one dimension or two dimensions. Tag bit “1″ indicates one-dimensional coding, and ” 0 ” indicates two – dimensional coding. The pattern of six consecutive synchronization codes (EOL plus tag bit) is called RTC in the MR coding scheme. In MR, each line of data is compared with the previous line for making code changes. After a line has been coded, it becomes the reference line for the next coding line. This coding scheme is likely to create errors in the picture data. A limit exists on how many lines of data are compared with a reference line. If the reference line has an error, then the same error is reproduced on any line that was compared with it. To avoid transmission errors at the end of the page, a one-dimensional code line is sent for every second or fourth line depending on the scanning density.
Transmission Time Per Total Coded Scan Line. In group-1 and group-2 modes, the minimum scan line time was fixed. Because of coding in G3 and G4 modes, the amount of compressed data varies from line to line. Therefore, the minimum scan line time (MSLT) supported by fax machines must indicate to each other before actual communication takes place. MSLT is the minimum time required by the receiving device for printing or buffering an image line. During the initial handshake, the receiving fax machine sets the digital identification signal (DIS) command for minimum scan line time. This information
is communicated to the sending device to ensure that image data are not transmitted faster than the receiving device can process it. The sending device will pad the image line with zeros to increase the transmission time to match the receiving device requirements. The total coded scan line is defined as the sum of data bits, fills bits, and EOL bits. For the optional two-dimensional coding schemes, the total coded scan line is given as a sum of data bits, required fill bits, and EOL bits plus a tag bit. The minimum transmission time of the total coded scan line is defined as the total time taken for transmission of the total coded scan line that is 20ms [ITU-T-T.4 (2003)]. These values vary with image data in the page and resolution. Based on printing methods, there are several optimal minimum transmission times of total coded scan line in addition to the standard 20 ms.
14.2.3

Modified Modified Read (MMR) Scheme

MMR is the basic facsimile coding scheme used in G4 ISDN facsimile devices. For G4, MMR is used for better compression than the MR scheme. The MMR coding is also known as the extended two-dimensional coding scheme that is described in T.6 [ITU-T-T.6 (1988)]. This coding scheme became a recommendation in 1992 and is used in G3 fax machines. The MMR coding scheme works similarly to the modified read scheme, but only two-dimensional lines are transmitted without fill bits and without EOL codes for synchronization. The coding scheme uses a two-dimensional line-by-line coding. The first reference line is assumed as a white line. The line to be coded is called the coding line, and it is coded with reference to the position of the elements of the reference line. After coding, the coding line becomes the reference line for the next line of data. An error-free communication link in ISDN makes it possible to transmit pages without errors. This coding scheme has to be used with the ECM option.
14.2.4

JPEG Image Coding

Continuous-tone color and grayscale modes are optional features of G3, which enable transmission of color or grayscale images. Color and grayscale modes were incorporated into G3 in 1994 [ITU-T-T.42 (1996)] by introducing JPEG image compression. This compression is explored in the T.81 recommendation [ITU-T-T.81 (1993)]. JPEG provides a compression method that is capable of compressing continuous-tone image data with a pixel depth of 6 to 24 bits with reasonable speed and efficiency. JPEG is popularly used in still pictures compression. The JPEG specifies two classes of coding processes: loss and lossless. The loss compression processes are based on the discrete cosine transform (DCT), and the lossless are based on predictive techniques. Four modes of operation exist under which the various processes are defined, namely the sequential DCT, progressive DCT, sequential lossless, and the hierarchical mode. JPEG was not an efficient solution to transmit the text or line art. JPEG
can compress the text and line art, but to get good compression efficiency, image quality suffers. JBIG is more suitable for fax images that code efficiently. JBIG gained more popularity than JPEG for fax images.
14.2.5

JBIG Coding

The JBIG group is a recommendations committee that had its origins within the International Standards Organization (ISO) and is a method for compressing efficiently bilevel, two-color, multilevel gray, and color images data. The main features of JBIG are given here:
• Lossless compression of one-bit-per-pixel image data
• Ability to encode individual bit planes of multiple bit pixels, catering to gray – level and color images
• Progressive or sequential encoding of image data
• More efficient and scalable for different resolutions
Most information transmitted through a facsimile device contains only black-and-white bilevel information. Most infrastructure of the transmission and the printing system are binary, and the output quality of this type is acceptable. JBIG addresses efficient coding of both bilevel and multilevel gray and color images as defined in T.82 [ITU-T-T.82 (1993)]. JBIG is intended to replace the less efficient MR and MMR compression algorithms used by the G3 and G4 fax devices.

Next post:

Previous post: