ZAR image format

Author:	ABel [TeamX] Last revision:	16 Jan 2001

This document is an attempt to describe the structure of ZAR images (image files in Fallout Tactics: Brotherhood of Steel game).

The ZAR format is the property of Interplay as stated in the "Fallout Tactics" readme file.

Note: this document introduces some terms, which are quoted through the document. The author does not pretend these terms are accurate or adequate.

One more note: the purpose of several fields is unknown, so some suppositions on their meaning are made, but neither can they be proved nor rejected by the author at the time of writing the document. Such hypotheses are marked with '(???)'.

I. Main features
1. The zar-image can be "stand-alone" and "embedded". I.e., it can be a single file (*.zar) or a part of some larger file (included into a Tile (*.til) or a Sprite (*.spr)). Class of "Embedded" ZARs can be subdivided into "complete" or "incomplete".

"Complete" ZAR contains all the information necessary for its displaying, though the enclosing file can provide some additional fields changing the way the image is showed.

"Incomplete" images do not have they own palettes (such ZARs can be observed in the Sprites). The color information is taken from the enclosing entity's palette.

2. The format uses the palette to store the color information. The palette entries are RGBA quadruples, besides the usual 256-level RGB there is an alpha- (or transparency-) channel, which is used to display shadows or anti-aliased images against any background. The alpha is also 256-leveled, where 255 stands for full opacity and 0 means full transparency.

3. The format uses a kind of RLE compression (though the term RLE is not quite adequate here). Pixel information is stored in blocks consiting of a header and pixel data bytes (optionally).

II. Format specification
This specification uses C-like constructions and expressions along with two self-explanatory pseudo-keywords: 'optional' and 'variable_length'. The following types are used:

CHAR	one byte integer (unsigned char in C) DWORD	four bytes integer (unsigned long in C). ZAR_IMAGE ::= variable_length struct { ZAR_HEADER image_header; optional ZAR_PALETTE palette; DWORD data_length; variable_length union { CHAR image_data[data_length]; RLE_BLOCK blocks[]; } }

ZAR_HEADER ::= struct { CHAR signature[6]; // zar-image signature, must be ' \0' CHAR zar_type; // must be 0x33 or 0x34 CHAR dummy1; // nothing (???) DWORD image_width, image_heigth; CHAR palette_present_flag; }

If palette_present_flag is 0x01 then zar_image includes a palette.

ZAR_PALETTE ::= variable_length struct { DWORD color_count; // count of color entries in the palette; // less than or equals to 256 (???) PAL_ENTRY colors[color_count]; // color entries optional CHAR default_color; // index of default (or background) // color (???); the field is   // included only when the zar_type // is 0x34. } PAL_ENTRY ::= struct { CHAR blue, green, red, alpha; } RLE_BLOCK ::= variable_length struct { CHAR rle_command : 2; // command (two bits) CHAR block_length : 6; // data length (six bits) CHAR pixel_data[]; // pixel information; the length of this // item can be zero, block_length or   // twice the block_length, depending on    // the value of rle_command. }

III. Decompression scheme
Pixel data is RLEncoded in the blocks of type RLE_BLOCK (or in the image_data array - they are the same things). The decompression starts from the top left corner of the image and stops when all the image_data array is processed or when the all the image_height scan lines are rendered, whichever happens first. The processing is carried out from left to right in the horizontal direction. On reaching of the right border (image_width field) the processing is switched to the start of the next scanline of the image.

Every block determines the color of the block_length consecutive pixels. The following table describes the treatment of pixel_data array and the color values of the next block_length image pixels depending on the rle_command used in a block.

rle_command | pixel_data          | image pixels -+--+---  0           | no data at all;      | next block_length pixels are filled | zero length         | with 00:00:00:00 (R:G:B:A) color, or               |                      | simply skipped -+--+---  1           | CHAR color_inds[b_l] | pixel color (RGB) is taken from the |                     | respective palette entries, the |                     | alpha-value (A) is set to 0xFF -+--+---  2           | { CHAR color_ind;    | pixel's RGBA is specified by the pair |  CHAR alpha_val;}   | from pixel_data array | pairs[b_l]          | -+--+---  3           | CHAR alpha_vals[b_l] | pixel's color (RGB) is set to the RGB |                     | value from colors[0] or may be from |                     | colors[default_color] (???), |                     | the alpha value is taken from the |                     | pixel_data array

(???) The block cannot span multiple scanlines. If the block defines more pixels than can fit into the rest of current scanline, the redundant data is to be ignored. (???)

If you have any questions regarding this document, or have found errors, mistakes or inaccuracies, please do not hesitate to email me to abel@krasu.ru or a234238@chat.ru

Found in TeamX's offline docs