Apple SMC

From MultimediaWiki
Jump to navigation Jump to search

This page is based on the document 'Description of the Apple Graphics (SMC) Codec' by Mike Melanson found at http://multimedia.cx/smc.txt.

The Apple SMC codec is also known as the Apple Graphics codec. It is a vector quantizer video codec that operates on 4x4 blocks of 8-bit paletted pixel values. The codec is named after its developer, Sean M. Callahan.

Basics of SMC Data and Decoding

All multi-byte values (there is actually only 1) are stored in big-endian format.

An SMC decoder renders 4x4 pixel blocks from left to right, top to bottom. The source data are 8-bit paletted values. The palette for an SMC-encoded Quicktime file is transported within the stsd atom of the video header.

An SMC decoder maintains 3 circular caches consisting of 256 elements each. Each element contains 2, 4 or 8 colors which are used in decoding 2-, 4-, and 8-color blocks:

  • color pair cache for 2-color encoded blocks:
           +---------+---------+
    pair 0 | color 0 | color 1 |
           +---------+---------+
    pair 1 | color 0 | color 1 |
           +---------+---------+
           /   ..        ..    /
           +---------+---------+
  pair 255 | color 0 | color 1 |
           +---------+---------+
  • color quad cache for 4-color encoded blocks:
           +---------+---------+---------+---------+
    quad 0 | color 0 | color 1 | color 2 | color 3 |
           +---------+---------+---------+---------+
    quad 1 | color 0 | color 1 | color 2 | color 3 |
           +---------+---------+---------+---------+
           /   ..        ..        ..        ..    /
           +---------+---------+---------+---------+
  quad 255 | color 0 | color 1 | color 2 | color 3 |
           +---------+---------+---------+---------+
  • color octet cache for 8-color encoded blocks:
           +---------+---------+---------+---------+
   octet 0 | color 0 | color 1 /   ..    / color 7 |
           +---------+---------+---------+---------+
   octet 1 | color 0 | color 1 /   ..    / color 7 |
           +---------+---------+---------+---------+
           /   ..        ..        ..        ..    /
           +---------+---------+---------+---------+
 octet 255 | color 0 | color 1 |   ..    | color 7 |
           +---------+---------+---------+---------+

The caches operate as circular arrays: If a cache exceeds 256 elements, new elements are stored starting from the beginning of the cache. All 3 caches are reset before decoding a new frame.

The first byte of a chunk is probably a flags byte. XAnim's comments seem to imply that this byte is always 0xe1. I've also observed a 0x80 byte in this byte. In either case, the meaning of the byte is unknown.

Bytes 2-4 are the length of the chunk. This value should match the value transported in the Quicktime file's video chunk length atom. Then, the bytestream in the chunk is traversed. The decoder renders blocks based on the upper nibble of command opcodes.

SMC Opcodes

After the first 4 bytes, an SMC chunk is a stream of opcodes and associated data. The opcode is the upper nibble of the next byte in the stream. The meaning of each opcode is detailed next.

0x00 or 0x10: Skip Blocks

These 2 opcodes instruct the decoder to skip blocks in the output frame. If there are 4x4 pixel blocks in the previous frame that are the same as in the current frame, there's no point in rendering them again, so these opcodes allow the decoder to skip blocks.

These opcodes skip n blocks, where n is 1 plus the lower nibble of the opcode byte if the opcode is 0x00, or 1 plus the next byte in the stream if the opcode is 0x10.

0x20 or 0x30: Repeat Last Block

These 2 opcodes instruct the decoder to repeat the last block over the next n blocks, where n is 1 plus the lower nibble of the opcode byte if the opcode is 0x20, or 1 plus the next byte in the stream if the opcode is 0x30. If the first block to be rendered is the first block on a line, repeat the last block from the previous line.

0x40 or 0x50: Repeat Previous 2 Blocks

These 2 opcodes instruct the decoder to repeat the last pair blocks over the next n pairs of blocks, where n is 1 plus the lower nibble of the opcode byte if the opcode is 0x40, or 1 plus the next byte in the stream if the opcode is 0x50. If the first block to be rendered is the first block on a lines, repeat the last 2 blocks from the previous line. Similarly, if the first block to be rendered is the second block on a line, the first repeated block will be the last block from the previous line, while the second repeated block will be the first block from the current line.

0x60 or 0x70: 1-Color Encoding

These 2 opcodes instruct the decoder to paint the next n blocks the same color. The number of blocks is 1 plus the lower nibble of the opcode byte if the opcode is 0x60, or 1 plus the next byte in the stream if the opcode is 0x70. The color to paint the block is the palette entry indexed by the next byte in the stream.

0x80 or 0x90: 2-Color Encoding

These 2 codes both specify a pair of colors with which to paint the next n blocks. The number of blocks to paint is equal to 1 plus the lower nibble of the opcode.

The color pair to be used to paint the blocks depends on the opcode. If the opcode is 0x80, the next 2 bytes in the stream specify the palette indices to be used. The 2 colors are stored in the next available entry in the color pair cache. If the opcode is 0x90, then the next byte in the stream specifies an index into the color pair cache containing the color pair to be used.

For each block to be rendered, there are 2 bytes (called a and b in this example) in the stream that are arrays of flags which specify which of the 2 colors in the selected pair will paint which pixels in the block. The flags are arranged as follows:

 a7 a6 a5 a4
 a3 a2 a1 a0
 b7 b6 b5 b4
 b3 b2 b1 b0

For example, it bit 3 of byte b is 1, then the pixel in the lower left corner of the block is set to color 1 from the selected pair, otherwise it is set to color 0.

0xA0 or 0xB0: 4-Color Encoding

These 2 codes both specify a quad of colors with which to paint the next n blocks. The number of blocks to paint is equal to 1 plus the lower nibble of the opcode.

The color quad to be used to paint the blocks depends on the opcode. If the opcode is 0xA0, the next 4 bytes in the stream specify the palette indices to be used. The 4 colors are stored in the next available entry in the color quad cache. If the opcode is 0xB0, then the next byte in the stream specifies an index into the color quad cache containing the color quad to be used.

For each block to be rendered, there are 4 bytes (called a, b, c and d in this example) in the stream that are arrays of flags which specify which of the 4 colors in the selected pair will paint which pixels in the block. The flags are arranged as follows:

 a76 a54 a32 a10
 b76 b54 b32 b10
 c76 c54 c32 c10
 d76 d54 d32 d10

For example, if the lower 2 bits of byte a are 11 (3 decimal), then color 3 of the selected color quad is placed in the upper right corner of the block.

0xC0 or 0xD0: 8-Color Encoding

These 2 codes both specify an octet of colors with which to paint the next n blocks. The number of blocks to paint is equal to 1 plus the lower nibble of the opcode.

The color octet to be used to paint the blocks depends on the opcode. If the opcode is 0xC0, the next 8 bytes in the stream specify the palette indices to be used. The 8 colors are stored in the next available entry in the color octet cache. If the opcode is 0xD0, then the next byte in the stream specifies an index into the color octet cache containing the color octet to be used.

For each block to be rendered, there are 6 bytes in the stream that will comprise arrays of flags that specify which of the 8 colors in the selected pair will paint which pixels in the block. It's best to assemble the next 6 bytes into 2 24-bit numbers stored in 2 32-bit variables (a and b in this example). The next 6 bytes in the stream are rearranged in the most unusual manner. For 6 bytes:

 [01 23 45 67 89 AB]

broken down into a vector of 12 nibbles:

 [n0n1 n2n3 n4n5 n6n7 n8n9 nAnB]

this is the arrangement of color flags:

 flags_a = n0n1 n2n4 n5n6
 flags_b = n8n9 nAn3 n7nB

The flags are arranged as follows:

 a23-21 a20-18 a17-15 a14-12
 a11-9  a8-6   a5-3   a2-0
 b23-21 b20-18 b17-15 b14-12
 b11-9  b8-6   b5-3   b2-0

For example, bits 23-21 of a32 will yield the index into the selected color octet to specify the color of the first pixel of the block. If bits 23-21 of a32 are 101 (5 decimal), color 5 from the selected octet is placed in the first pixel of the block.

0xE0: 16-Color Encoding

For n blocks, where n is 1 plus the lower nibble of the command opcode (no exception here), copy the blocks directly into the output image. Each block is comprised of 16 bytes and each byte is a palette entry. For bytes [0, 1, .., 15] in the bytestream, the block is laid out as

  0  1  2  3
  4  5  6  7
  8  9 10 11
 12 13 14 15

0xF0: Unknown Opcode

The XAnim source code does not necessarily list this as an invalid opcode, just an unknown one.

References