Graphics Interchange Format -- Apendices

Quick Reference Table | Gif Grammer | Glossary | Conventions | Interlaced Images | LZW Compression | On-Line Capabilities

Appendix A. Quick Reference Table.

Block Name                  Required   Label       Ext.   Vers.
Application Extension       Opt. (*)   0xFF (255)  yes    89a
Comment Extension           Opt. (*)   0xFE (254)  yes    89a
Global Color Table          Opt. (1)   none        no     87a
Graphic Control Extension   Opt. (*)   0xF9 (249)  yes    89a
Header                      Req. (1)   none        no     N/A
Image Descriptor            Opt. (*)   0x2C (044)  no     87a (89a)
Local Color Table           Opt. (*)   none        no     87a
Logical Screen Descriptor   Req. (1)   none        no     87a (89a)
Plain Text Extension        Opt. (*)   0x01 (001)  yes    89a
Trailer                     Req. (1)   0x3B (059)  no     87a

Unlabeled Blocks
Header                      Req. (1)   none        no     N/A
Logical Screen Descriptor   Req. (1)   none        no     87a (89a)
Global Color Table          Opt. (1)   none        no     87a
Local Color Table           Opt. (*)   none        no     87a

Graphic-Rendering Blocks
Plain Text Extension        Opt. (*)   0x01 (001)  yes    89a
Image Descriptor            Opt. (*)   0x2C (044)  no     87a (89a)

Control Blocks
Graphic Control Extension   Opt. (*)   0xF9 (249)  yes    89a

Special Purpose Blocks
Trailer                     Req. (1)   0x3B (059)  no     87a
Comment Extension           Opt. (*)   0xFE (254)  yes    89a
Application Extension       Opt. (*)   0xFF (255)  yes    89a

(1)   if present, at most one occurrence
(*)   zero or more occurrences
(+)   one or more occurrences

Notes : The Header is not subject to Version Numbers. (89a) The Logical Screen Descriptor and the Image Descriptor retained their syntax from version 87a to version 89a, but some fields reserved under version 87a are used under version 89a.

Appendix B. GIF Grammar.

A Grammar is a form of notation to represent the sequence in which certain objects form larger objects. A grammar is also used to represent the number of objects that can occur at a given position. The grammar given here represents the sequence of blocks that form the GIF Data Stream. A grammar is given by listing its rules. Each rule consists of the left-hand side, followed by some form of equals sign, followed by the right-hand side. In a rule, the right-hand side describes how the left-hand side is defined. The right-hand side consists of a sequence of entities, with the possible presence of special symbols. The following legend defines the symbols used in this grammar for GIF.

Legend: <> grammar word ::= defines symbol * zero or more occurrences + one or more occurrences | alternate element [] optional element


<GIF Data Stream> ::= Header <Logical Screen> <Data>* Trailer

This rule defines the entity <GIF Data Stream> as follows. It must begin with a Header. The Header is followed by an entity called Logical Screen, which is defined below by another rule. The Logical Screen is followed by the entity Data, which is also defined below by another rule. Finally, the entity Data is followed by the Trailer. Since there is no rule defining the Header or the Trailer, this means that these blocks are defined in the document. The entity Data has a special symbol (*) following it which means that, at this position, the entity Data may be repeated any number of times, including 0 times. For further reading on this subject, refer to a standard text on Programming Languages.

The Grammar.

<GIF Data Stream> ::= Header <Logical Screen> <Data>* Trailer

<Logical Screen> ::= Logical Screen Descriptor [Global Color Table]

<Data> ::= <Graphic Block> | <Special-Purpose Block>

<Graphic Block> ::= [Graphic Control Extension] <Graphic-Rendering Block>

<Graphic-Rendering Block> ::= <Table-Based Image> | Plain Text Extension

<Table-Based Image> ::= Image Descriptor [Local Color Table] Image Data

<Special-Purpose Block> ::= Application Extension | Comment Extension

NOTE : The grammar indicates that it is possible for a GIF Data Stream to contain the Header, the Logical Screen Descriptor, a Global Color Table and the GIF Trailer. This special case is used to load a GIF decoder with a Global Color Table, in preparation for subsequent Data Streams without color tables at all.

Appendix C. Glossary.

Active Color Table - Color table used to render the next graphic. If the next graphic is an image which has a Local Color Table associated with it, the active color table becomes the Local Color Table associated with that image. If the next graphic is an image without a Local Color Table, or a Plain Text Extension, the active color table is the Global Color Table associated with the Data Stream, if there is one; if there is no Global Color Table in the Data Stream, the active color table is a color table saved from a previous Data Stream, or one supplied by the decoder.

Block - Collection of bytes forming a protocol unit. In general, the term includes labeled and unlabeled blocks, as well as Extensions.

Data Stream - The GIF Data Stream is composed of blocks and sub-blocks representing images and graphics, together with control information to render them on a display device. All control and data blocks in the Data Stream must follow the Header and must precede the Trailer.

Decoder - A program capable of processing a GIF Data Stream to render the images and graphics contained in it.

Encoder - A program capable of capturing and formatting image and graphic raster data, following the definitions of the Graphics Interchange Format.

Extension - A protocol block labeled by the Extension Introducer 0x21.

Extension Introducer - Label (0x21) defining an Extension.

Graphic - Data which can be rendered on the screen by virtue of some algorithm. The term graphic is more general than the term image; in addition to images, the term graphic also includes data such as text, which is rendered using character bit-maps.

Image - Data representing a picture or a drawing; an image is represented by an array of pixels called the raster of the image.

Raster - Array of pixel values representing an image.

Appendix D. Conventions.

Animation - The Graphics Interchange Format is not intended as a platform for animation, even though it can be done in a limited way.

Byte Ordering - Unless otherwise stated, multi-byte numeric fields are ordered with the Least Significant Byte first.

Color Indices - Color indices always refer to the active color table, either the Global Color Table or the Local Color Table.

Color Order - Unless otherwise stated, all triple-component RGB color values are specified in Red-Green-Blue order.

Color Tables - Both color tables, the Global and the Local, are optional; if present, the Global Color Table is to be used with every image in the Data Stream for which a Local Color Table is not given; if present, a Local Color Table overrides the Global Color Table. However, if neither color table is present, the application program is free to use an arbitrary color table. If the graphics in several Data Streams are related and all use the same color table, an encoder could place the color table as the Global Color Table in the first Data Stream and leave subsequent Data Streams without a Global Color Table or any Local Color Tables; in this way, the overhead for the table is eliminated. It is recommended that the decoder save the previous Global Color Table to be used with the Data Stream that follows, in case it does not contain either a Global Color Table or any Local Color Tables. In general, this allows the application program to use past color tables, significantly reducing transmission overhead.

Extension Blocks - Extensions are defined using the Extension Introducer code to mark the beginning of the block, followed by a block label, identifying the type of extension. Extension Codes are numbers in the range from 0x00 to 0xFF, inclusive. Special purpose extensions are transparent to the decoder and may be omitted when transmitting the Data Stream on-line. The GIF capabilities dialogue makes the provision for the receiver to request the transmission of all blocks; the default state in this regard is no transmission of Special purpose blocks.

Reserved Fields - All Reserved Fields are expected to have each bit set to zero (off).

Appendix E. Interlaced Images.

The rows of an Interlaced images are arranged in the following order:

Group 1 : Every 8th. row, starting with row 0. (Pass 1)

Group 2 : Every 8th. row, starting with row 4. (Pass 2)

Group 3 : Every 4th. row, starting with row 2. (Pass 3)

Group 4 : Every 2nd. row, starting with row 1. (Pass 4)

The Following example illustrates how the rows of an interlaced image are ordered.

Row Number                                        Interlace Pass

0    -----------------------------------------       1
1    -----------------------------------------       4
2    -----------------------------------------       3
3    -----------------------------------------       4
4    -----------------------------------------       2
5    -----------------------------------------       4
6    -----------------------------------------       3
7    -----------------------------------------       4
8    -----------------------------------------       1
9    -----------------------------------------       4
10   -----------------------------------------       3
11   -----------------------------------------       4
12   -----------------------------------------       2
13   -----------------------------------------       4
14   -----------------------------------------       3
15   -----------------------------------------       4
16   -----------------------------------------       1
17   -----------------------------------------       4
18   -----------------------------------------       3
19   -----------------------------------------       4

Appendix F. Variable-Length-Code LZW Compression.

The Variable-Length-Code LZW Compression is a variation of the Lempel-Ziv Compression algorithm in which variable-length codes are used to replace patterns detected in the original data. The algorithm uses a code or translation table constructed from the patterns encountered in the original data; each new pattern is entered into the table and its index is used to replace it in the compressed stream.

The compressor takes the data from the input stream and builds a code or translation table with the patterns as it encounters them; each new pattern is entered into the code table and its index is added to the output stream; when a pattern is encountered which had been detected since the last code table refresh, its index from the code table is put on the output stream, thus achieving the data compression. The expander takes input from the compressed data stream and builds the code or translation table from it; as the compressed data stream is processed, codes are used to index into the code table and the corresponding data is put on the decompressed output stream, thus achieving data decompression. The details of the algorithm are explained below. The Variable-Length-Code aspect of the algorithm is based on an initial code size (LZW-initial code size), which specifies the initial number of bits used for the compression codes. When the number of patterns detected by the compressor in the input stream exceeds the number of patterns encodable with the current number of bits, the number of bits per LZW code is increased by one.

The Raster Data stream that represents the actual output image can be represented as:

    7 6 5 4 3 2 1 0
   | LZW code size |

   +---------------+ ----+
   |  block size   |     |
   +---------------+     |
   |               |     +-- Repeated as many
   |  data bytes   |     |   times as necessary.
   |               |     |
   +---------------+ ----+
   . . . Terminate Code for compressed data must appear before Block Terminator
   |0 0 0 0 0 0 0 0|  Block Terminator

The conversion of the image from a series of pixel values to a transmitted or stored character stream involves several steps. In brief these steps are:

  1. Establish the Code Size - Define the number of bits needed to represent the actual data.

  2. Compress the Data - Compress the series of image pixels to a series of

    compression codes.

  3. Build a Series of Bytes - Take the set of compression codes and convert to a string of 8-bit bytes.

  4. Package the Bytes - Package sets of bytes into blocks preceded by character counts and output.

Establish Code Size

The first byte of the Compressed Data stream is a value indicating the minimum number of bits required to represent the set of actual pixel values. Normally this will be the same as the number of color bits. Because of some algorithmic constraints however, black & white images which have one color bit must be indicated as having a code size of 2. This code size value also implies that the compression codes must start out one bit longer.


The LZW algorithm converts a series of data values into a series of codes which may be raw values or a code designating a series of values. Using text characters as an analogy, the output code consists of a character or a code representing a string of characters.

The LZW algorithm used in GIF matches algorithmically with the standard LZW algorithm with the following differences:

1. A special Clear code is defined which resets all compression/decompression parameters and tables to a start-up state. The value of this code is 2**<code size>. For example if the code size indicated was 4 (image was 4 bits/pixel) the Clear code value would be 16 (10000 binary). The Clear code can appear at any point in the image data stream and therefore requires the LZW algorithm to process succeeding codes as if a new data stream was starting. Encoders should output a Clear code as the first code of each image data stream.

2. An End of Information code is defined that explicitly indicates the end of the image data stream. LZW processing terminates when this code is encountered. It must be the last code output by the encoder for an image. The value of this code is +1.

3. The first available compression code value is <clear code>+2.

4. The output codes are of variable length, starting at <code size>+1 bits per code, up to 12 bits per code. This defines a maximum code value of 4095 (0xFFF). Whenever the LZW code value would exceed the current code length, the code length is increased by one. The packing/unpacking of these codes must then be altered to reflect the new code length.

Build 8-Bit Bytes

Because the LZW compression used for GIF creates a series of variable length codes, of between 3 and 12 bits each, these codes must be reformed into a series of 8-bit bytes that will be the characters actually stored or transmitted. This provides additional compression of the image. The codes are formed into a stream of bits as if they were packed right to left and then picked off 8 bits at a time to be output.

Assuming a character array of 8 bits per character and using 5 bit codes to be packed, an example layout would be similar to:

0  |               |    bbbaaaaa
1  |               |    dcccccbb
2  |               |    eeeedddd
3  |               |    ggfffffe
4  |               |    hhhhhggg
   . . .
N  |               |

Note that the physical packing arrangement will change as the number of bits per compression code change but the concept remains the same.

Package The Bytes

Once the bytes have been created, they are grouped into blocks for output by preceding each block of 0 to 255 bytes with a character count byte. A block with a zero byte count terminates the Raster Data stream for a given image. These blocks are what are actually output for the GIF image. This block format has the side effect of allowing a decoding program the ability to read past the actual image data if necessary by reading block counts and then skipping over the data.

Further Reading

  1. Ziv, J. and Lempel, A. : "A Universal Algorithm for Sequential Data Compression", IEEE Transactions on Information Theory, May 1977.
  2. Welch, T. : "A Technique for High-Performance Data Compression", Computer, June 1984.
  3. Nelson, M.R. : "LZW Data Compression", Dr. Dobb's Journal, October 1989.

Appendix G. On-line Capabilities Dialogue.

NOTE : This section is currently (10 July 1990) under revision; the information provided here should be used as general guidelines. Code written based on this information should be designed in a flexible way to accommodate any changes resulting from the revisions.

The following sequences are defined for use in mediating control between a GIF sender and GIF receiver over an interactive communications line. These sequences do not apply to applications that involve downloading of static GIF files and are not considered part of a GIF file.

Gif Capabilities Enquiry

The GIF Capabilities Enquiry sequence is issued from a host and requests an interactive GIF decoder to return a response message that defines the graphics parameters for the decoder. This involves returning information about available screen sizes, number of bits/color supported and the amount of color detail supported. The escape sequence for the GIF Capabilities Enquiry is defined as:

ESC[>0g 0x1B 0x5B 0x3E 0x30 0x67

Gif Capabilities Response

The GIF Capabilities Response message is returned by an interactive GIF decoder and defines the decoder's display capabilities for all graphics modes that are supported by the software. Note that this can also include graphics printers as well as a monitor screen. The general format of this message is:

#version;protocol{;dev, width, height, color-bits, color-res}...

'#' GIF Capabilities Response identifier character. version GIF format version number; initially '87a'. protocol='0' No end-to-end protocol supported by decoder Transfer as direct 8-bit data stream. protocol='1' Can use CIS B+ error correction protocol to transfer GIF data interactively from the host directly to the display. dev = '0' Screen parameter set follows. dev = '1' Printer parameter set follows. width Maximum supported display width in pixels. height Maximum supported display height in pixels. color-bits Number of bits per pixel supported. The number of supported colors is therefore 2**color-bits. color-res Number of bits per color component supported in the hardware color palette. If color-res is '0' then no hardware palette table is available.

Note that all values in the GIF Capabilities Response are returned as ASCII decimal numbers and the message is terminated by a Carriage Return character.

The following GIF Capabilities Response message describes three standard IBM PC Enhanced Graphics Adapter cponfigurations with no printer; the GIF data stream can be processed within an error correcting protocol:


Enter Gif Graphics Mode

Two sequences are currently defined to invoke an interactive GIF decoder into action. The only difference between them is that different output media are selected. These sequences are:

ESC[>1g Display GIF image on screen

0x1B 0x5B 0x3E 0x31 0x67

ESC[>2g Display image directly to an attached graphics printer. The image may optionally be displayed on the screen as well.

0x1B 0x5B 0x3E 0x32 0x67

Note that the 'g' character terminating each sequence is in lowercase.

Interactive Environment

The assumed environment for the transmission of GIF image data from an interactive application is a full 8-bit data stream from host to micro. All 256 character codes must be transferrable. The establishing of an 8-bit data path for communications will normally be taken care of by the host application programs. It is however up to the receiving communications programs supporting GIF to be able to receive and pass on all 256 8-bit codes to the GIF decoder software.

Free Spam Protection   Android ORM   Simple Java Magic   JMX using HTTP   Great Eggnog Recipe   Massachusetts Covid Vaccine Sites