Kostya's Boring Codec World

I’ve been reminded that this variant of TrueMotion exists too. What do you know, it’s actually somewhat like TrueMotion 2 NoModifiers.

Essentially it’s just another fixed packing scheme like Creative YUV, Cirrus Logic CLJR or Aura. You have left prediction, deltas coded with nibbles, the usual stuff (at least blocks in TM2 were coded similarly). The only peculiar thing is that it codes data by planes with chroma planes being coded first.

I hope to add detailed description of this codec to Multimedia Wiki by the end of this week and then forget about it again.

So, while I still have no idea how this codecs functions I can describe some technical details from it.

First, codec data consists of chunks with tag like 0xA00001xx and chunk size in the beginning. Some chunks are unique, some may repeat, some are alternative to each other (e.g. there are four different chunk IDs for Huffman tree description, two of them differ only in header before tree data).

Second, some smaller chunks (like 0x09 with 3-byte payload containing some decoding parameters) are obfuscated by XORing with the key derived from main chunk data. Annoying and not adding much protection really.

Third, unlike plain TrueMotion 2, TrueMotion 2X11R6 has 8×8 blocks (and not 4×4), only 3 block types (instead of 7) and single Huffman tree descriptor (instead of one per non-null block types plus one for block types itself). And it’s in a rather curious format too.

Typical TM2X (or TM2A) frame usually (i.e. for both known samples) consists of 0x06 chunk with compressed block data, some small chunks like 0x15, 0x09, two 0x02 chunks, about a dozen of 0x0B chunks and 0x0A chunk with Huffman code description.

Motion vector coding is represented in several variations: simple signed 8-bit values, MV vector of fixed bit size with bias (both are coded before MV data), some recursive MV coding for large frame areas and even the coding using Huffman coding.

And finally some notes about Huffman coding itself. I’ve not understood it properly yet but here are some notes:

• Huffman code descriptor is actually a 2D table of 8×256 size (it’s stored in compact way in the corresponding chunk), i.e. every byte has a list of up to 8 elements corresponding to it;
• decoding is performed by moving on the 8-element list unless an escape value is seen, then a byte is read from the input and new 8-element list is selected, and after decoding the current position is saved for later (e.g. first you read byte like 0x2A and it corresponds to a list 0, 1, 2, 0x83 — that means on subsequent decoding calls you should get 0, 1, 2, 3 and move to reading a new byte from input). Disclaimer: at least that’s how I understood it, it seems to be a reverse coding to me, i.e. assigning a variable amount of tokens to single byte of input instead of conventional assigning a variable amount of bits to the single token;
• in some cases an additional value may be read using both the descriptor and some additional table (it’s added to the result in those cases).

Funny how I started this blog more than 10 years ago mostly to talk about TrueMotion 2 and now it’s TrueMotion 2X time.

First of all, an existing binary specification (feel free to ask Baidu for some other materials for this codec, I’m pretty sure you’ll receive a lot) is weird and half of it is not well decompilable. It looks like the compiler did something inverse to inlining and split out some parts of code into separate functions without usual prologue and simply accesses variables somewhere deep on stack:

(more…)

There are a lot of variants of first two TrueMotion codecs. Final Fantasy (some number here) for PC uses TM2.

Differences from reference decoder: delta table size must be greater (64 instead of 32) and sometimes there is no data in stream, which means you should fill it with value from Huffman table).

Well, there is also TM2X awaiting.

My TrueMotion 2 decoder is almost complete. It just has some problems with chroma decoding in intraframes. I suspect this is because of some rounding errors and such. Original TM2 decoder used to store chroma in pairs – one 4 double word for 16-bit Cr and Cb and all calculations were done on those pairs.

For now, I’ll leave this decoder for some time and will post pieces of information about other codecs.

TM2 frame is divided into blocks 4×4 and YUV420 colorspace is used, so for each 4×4 blocks of luminance there is two 2×2 blocks of chroma.
There are 7 types of blocks:

• Hi-res block (hi-res luma and chroma)
• Med-res block (hi-res luma and low-res chroma)
• Low res block (low-res chroma and luma)
• Null-res block (nothing changes from previous block)
• Still block (just copied from previous frame)
• Update block (copied from the same place in previous frame and then modified)
• Motion block

Every intraframe block (first four types in the list) uses such logic:

LAST 4 ELEMENTS
| | | |
V V V V
D0 -> +d00 +d10 +d20 +d30 -> D0
D1 -> +d01 +d11 +d21 +d31 -> D1
D2 -> +d02 +d12 +d22 +d32 -> D2
D3 -> +d03 +d13 +d23 +d33 -> D2
| | | |
V V V V
LAST 4 ELEMENTS

It means that every block use last 4 pixels from top block as reference and deltas D (differences between right pixel in this line and previous line) from left neighbour.
In case of low-res chroma or luma last elements and deltas are modified before processing, they are replaced with their average values.
For interframe blocks there is one additional operation – to calculate new deltas and last values from block data after opying is done.

TM2 differs from many modern and not modern codecs in aspect it uses different streams for different type of data – look at the scheme below:


[Global Header]
[Stream 0 - High-res chroma values]
[Stream 1 - Low-res chroma values]
[Stream 2 - High-res luminance values]
[Stream 3 - Low-res luminance values]
[Stream 4 - Update deltas values]
[Stream 5 - Motion vectors]
[Stream 6 - Block types]


Global header can be one of two types: old header type – all values are stored in 4-byte or 2-byte integers and new header type – header fields are stored in bitstream (9 bits for width and the same for height, 31 bits for flags, etc.).
One thing worth to mention – a mixture of big- and little-endian 4-byte integers in header values and use big-endian dwords in bitstream output.

Streams are huffman-coded array of values (called ‘tokens’) and consist of header, non-compulsory delta table (in this case tokens are just indices in that table), compressed Huffman tree and compressed tokens. Stream also can be empty (in case of intra-frames – that streams 4 and 5).

Decoder must decompress all these streams into memory and read tokens from one or another stream. Decoding process will be described in Part 2.

Duck Corp. (now On2) released GPLed source of their solution VPVision (capturing software for Windows), and it contains sources of their two codecs – TrueMotion 1 and TrueMotion 2 (predecessors of VP3-7 family).
TrueMotion 1 decoder is already present in FFmpeg, now it’s time to add TrueMotion 2.
The bad thing is that codec source code is poorly-documented and optimised for performance (optimisation is a form of obfuscation and vice-versa – look at The International Obfuscated C Code Contest for examples).