Bink Video - Revision history

Kostya at 09:26, 27 February 2010

2010-02-27T09:26:04Z

← Older revision		Revision as of 02:26, 27 February 2010
Line 1:		Line 1:
	* FourCC: ~~BINK~~ (note that some FourCC lists assert that BINK is a FourCC for general-purpose multimedia containers; however, Bink data is always known to be encapsulated in a custom container format)		* FourCC: BIKf, BIKg, BIKh, BIKi (note that some FourCC lists assert that BINK is a FourCC for general-purpose multimedia containers; however, Bink data is always known to be encapsulated in a custom container format)
	* Company: [[RAD Game Tools]]		* Company: [[RAD Game Tools]]
	* Samples: [http://samples.mplayerhq.hu/game-formats/bink/ http://samples.mplayerhq.hu/game-formats/bink/], countless video games		* Samples: [http://samples.mplayerhq.hu/game-formats/bink/ http://samples.mplayerhq.hu/game-formats/bink/], countless video games
Line 5:		Line 5:
	Bink is a video codec that uses [[Vector Quantization]] along with [[DCT]] and [[Motion Compensation]]). This codec is used in a large number of computer and console games. The video is packaged in custom [[Bink Container\|Bink]] files.		Bink is a video codec that uses [[Vector Quantization]] along with [[DCT]] and [[Motion Compensation]]). This codec is used in a large number of computer and console games. The video is packaged in custom [[Bink Container\|Bink]] files.

	Each plane is coded separately. ~~The Y~~ plane is ~~always coded first~~ in ~~each frame~~.		Each plane is coded separately. First plane is luma or alpha plane followed by luma plane, since version 'i' there is 32-bit word containing plane data size in bytes before alpha and luma plane, so they may be skipped or decoded in parallel with chroma.

	== Format ==		== Format ==

Kostya at 11:12, 3 September 2009

2009-09-03T11:12:41Z

← Older revision		Revision as of 04:12, 3 September 2009
Line 3:		Line 3:
	* Samples: [http://samples.mplayerhq.hu/game-formats/bink/ http://samples.mplayerhq.hu/game-formats/bink/], countless video games		* Samples: [http://samples.mplayerhq.hu/game-formats/bink/ http://samples.mplayerhq.hu/game-formats/bink/], countless video games

	Bink is a video codec that ~~purports to use every video coding technique (~~[[~~DCT~~]], [[~~FFT~~]]~~, [[Wavelet Coding]], [[Vector Quantization]],~~ [[Motion Compensation]]) ~~under the sun and~~ used in a large number of computer and console games. The video is packaged in custom [[Bink Container\|Bink]] files.		Bink is a video codec that uses [[Vector Quantization]] along with [[DCT]] and [[Motion Compensation]]). This codec is used in a large number of computer and console games. The video is packaged in custom [[Bink Container\|Bink]] files.

	Each plane is coded separately. The Y plane is always coded first in each frame.		Each plane is coded separately. The Y plane is always coded first in each frame.
Line 194:		Line 194:

	=== Coefficients reading ===		=== Coefficients reading ===
			==== 'lossy' block (residue) ====
			TODO

			==== 'lossless' block (DCT coeffs) ====
			TODO

	== Decoding ==		== Decoding ==
Line 205:		Line 210:
	* x motion offset - 4 bit signed bundle		* x motion offset - 4 bit signed bundle
	* y motion offset - 4 bit signed bundle		* y motion offset - 4 bit signed bundle
	* DC for ~~interblock?~~ - 16 bit delta		* DC for intrablock - 16 bit delta
	* DC for ~~intrablock?~~ - 16 bit delta		* DC for interblock - 16 bit delta
	* pattern colors run length - 4 bit bundle		* pattern colors run length - 4 bit bundle

	=== Block types ===		=== Block types ===

	~~'''NOTE: This section is incomplete'''~~

	==== 0 - skip block ====		==== 0 - skip block ====
Line 226:		Line 229:
	4 bits - pattern index		4 bits - pattern index

	first 63 coefficients in block are decoded in this way:		first 63 or 64 coefficients in block are decoded in this way:

	run = next value from Pattern Colors Run Length;		run = next value from Pattern Colors Run Length;
Line 238:		Line 241:
	block += run;		block += run;

	Last block element is taken from <code>Colors</code> values		Last block element (if decoded 63 coefficients) is taken from <code>Colors</code> values

	==== 4 - motion+residue ====		==== 4 - motion+residue ====
	~~looks like block type 7 without DCT, i~~.e. ~~MC and coded difference only~~, 7 bits - number of ~~elements coded~~		motion compensation + so-called "lossy" coding (see [http://codecs.multimedia.cx/?p=189]), 7 bits before coded block contain number of masks to decode.

	==== 5 - intra DCT ====		==== 5 - intra DCT ====
Line 256:		Line 259:

	==== 9 - raw data ====		==== 9 - raw data ====
	Fill 8x8 block with 64 values from <code>Colors</code		Fill 8x8 block with 64 values from <code>Colors</code>

	~~[[Category:Undiscovered Video Codecs]]~~
	[[Category:Video Codecs]]		[[Category:Video Codecs]]
	~~[[Category:Undiscovered Game Formats]]~~
	[[Category:Game Formats]]		[[Category:Game Formats]]

Kostya: /* Decoding */

2009-07-15T16:11:35Z

Decoding

← Older revision		Revision as of 09:11, 15 July 2009
Line 213:		Line 213:
	'''NOTE: This section is incomplete'''		'''NOTE: This section is incomplete'''

	=== 0 ===		==== 0 - skip block ====
	Copy 8x8 block from previous frame to current frame		Copy 8x8 block from previous frame to current frame

	=== 1 ===		==== 1 - scaled 16x16 block ====
	This is 16x16 block which has its own subtype, which corresponds to other block types (range 3-9). Blocks should be decoded in the same way and then scaled twice.		This is 16x16 block which has its own subtype, which corresponds to other block types (range 3-9). Blocks should be decoded in the same way and then scaled twice.

	=== 2 ===		==== 2 - motion block ====
	Copy 8x8 block from previous frame with some offset (x and y values decoded at the beginning of the block)		Copy 8x8 block from previous frame with some offset (x and y values decoded at the beginning of the block)

	=== 3 ===		==== 3 - run fill ====
	Fill 8x8 block with pattern		Fill 8x8 block with pattern
	4 bits - pattern index		4 bits - pattern index
Line 240:		Line 240:
	Last block element is taken from <code>Colors</code> values		Last block element is taken from <code>Colors</code> values

	=== 4 ===		==== 4 - motion+residue ====
	looks like block type 7 without DCT, i.e. MC and coded difference only, 7 bits - number of elements coded		looks like block type 7 without DCT, i.e. MC and coded difference only, 7 bits - number of elements coded

	=== 5 ===		==== 5 - intra DCT ====
	intrablock with DCT coefficients and 4 bits quantizer		intrablock with DCT coefficients and 4 bits quantizer

	=== 6 ===		==== 6 - fill ====
	Fill 8x8 block with one value from <code>Colors</code>		Fill 8x8 block with one value from <code>Colors</code>

	=== 7 ===		==== 7 - inter DCT ====
	interblock with motion vector (x and y values decoded at the beginning), DCT coefficients and 4 bit quantizer		interblock with motion vector (x and y values decoded at the beginning), DCT coefficients and 4 bit quantizer

	=== 8 ===		==== 8 - pattern fill ====
	Fill block with two colours taken from <code>Colors</code> values using 8 values from <code>Patterns</code>		Fill block with two colours taken from <code>Colors</code> values using 8 values from <code>Patterns</code>

	=== 9 ===		==== 9 - raw data ====
	Fill 8x8 block with 64 values from <code>Colors</code		Fill 8x8 block with 64 values from <code>Colors</code

Kostya: /* Bundle reading */

2009-07-11T05:27:08Z

Bundle reading

← Older revision		Revision as of 22:27, 10 July 2009
Line 86:		Line 86:
	return 0; //no data to decode		return 0; //no data to decode
	if(getbit())		if(getbit())
	memset(dst, ~~gethuff~~(~~bundle->tree[0]~~), t);		memset(dst, getbits(4), t);
	else{		else{
	for(i = 0; i < t; i++)		for(i = 0; i < t; i++)
Line 167:		Line 167:
	}		}

	==== 16 bit delta ~~bundle~~ ====		==== 16 bit delta ====
			This one does not employ Huffman coding and uses delta coding instead.

	t = getbits(bundle->runsize);		t = getbits(bundle->runsize);

Kostya: /* Bundle reading */

2009-07-08T10:32:43Z

Bundle reading

← Older revision		Revision as of 03:32, 8 July 2009
Line 191:		Line 191:
	}		}
	}		}

			=== Coefficients reading ===

	== Decoding ==		== Decoding ==

Kostya: /* Bundle reading */

2009-07-08T10:31:30Z

Bundle reading

@@ Line 77: / Line 77: @@
 === Bundle reading ===
--bit values are decoded with bundles. Bundle consists of 17 Huffman trees - one to decode upper 4 bits and 16 trees that are used to decode lower 4 bits. Those lower 4 bits are used as tree number to decode next low nibble.
+Bink uses basic Huffman trees for decoding 4-bit values to create values of different types from them.
 == Decoding ==

Kostya: /* Frame */

2009-07-08T08:20:39Z

Frame

← Older revision		Revision as of 01:20, 8 July 2009
Line 86:		Line 86:
	* sub block type - 4 bit RLE bundle		* sub block type - 4 bit RLE bundle
	* colors data (for filling blocks) - 8 bit bundle		* colors data (for filling blocks) - 8 bit bundle
	* pattern - 8 bit pair bundle		* pattern - 4 bit pair bundle
	* x motion offset - 4 bit signed bundle		* x motion offset - 4 bit signed bundle
	* y motion offset - 4 bit signed bundle		* y motion offset - 4 bit signed bundle

Kostya: /* Decoding */

2009-07-08T08:02:34Z

Decoding

← Older revision		Revision as of 01:02, 8 July 2009
Line 80:		Line 80:

	== Decoding ==		== Decoding ==
			=== Frame ===

			for each block a lot of values should be decoded (value may be taken from cache if previous decode returned run of values):
			* block type - 4 bit RLE bundle
			* sub block type - 4 bit RLE bundle
			* colors data (for filling blocks) - 8 bit bundle
			* pattern - 8 bit pair bundle
			* x motion offset - 4 bit signed bundle
			* y motion offset - 4 bit signed bundle
			* DC for interblock? - 16 bit delta
			* DC for intrablock? - 16 bit delta
			* pattern colors run length - 4 bit bundle

	=== Block types ===		=== Block types ===

Line 91:		Line 104:

	=== 2 ===		=== 2 ===
	Copy 8x8 block from previous frame with some offset (~~previous motion vector?~~)		Copy 8x8 block from previous frame with some offset (x and y values decoded at the beginning of the block)

	=== 3 ===		=== 3 ===
	Fill 8x8 block with pattern		Fill 8x8 block with pattern
	4 bits - pattern index		4 bits - pattern index
	63 ~~bits - pattern~~
			first 63 coefficients in block are decoded in this way:

			run = next value from Pattern Colors Run Length;
			if(getbit()){
			memset(block, Colors[0], run);
			Colors++;
			}else{
			memcpy(block, Colors, run);
			Colors += run;
			}
			block += run;

			Last block element is taken from <code>Colors</code> values

	=== 4 ===		=== 4 ===
	~~"lossy"~~-coded ~~DCT coefficients with motion vector~~		looks like block type 7 without DCT, i.e. MC and coded difference only, 7 bits - number of elements coded

	=== 5 ===		=== 5 ===
	~~"losslessly"-coded~~ DCT coefficients and 4 bits quantizer		intrablock with DCT coefficients and 4 bits quantizer

	=== 6 ===		=== 6 ===
	Fill 8x8 block with one ~~byte~~		Fill 8x8 block with one value from <code>Colors</code>

	=== 7 ===		=== 7 ===
	~~"losslessly"-coded~~ DCT coefficients and 4 ~~bits~~ quantizer ~~and "motion"(in DCT code)~~		interblock with motion vector (x and y values decoded at the beginning), DCT coefficients and 4 bit quantizer

	=== 8 ===		=== 8 ===
	Fill block using ~~masks~~		Fill block with two colours taken from <code>Colors</code> values using 8 values from <code>Patterns</code>

	=== 9 ===		=== 9 ===
	~~Copy~~ 8x8 block from ~~some other frame?~~		Fill 8x8 block with 64 values from <code>Colors</code

	[[Category:Undiscovered Video Codecs]]		[[Category:Undiscovered Video Codecs]]

Kostya: /* 1 */

2009-06-29T14:34:31Z

← Older revision		Revision as of 07:34, 29 June 2009
Line 88:		Line 88:

	=== 1 ===		=== 1 ===
	~~Unlike other blocks, this seem to work with~~ 16x16 block~~. It~~ has ~~a subcode for different real~~ block types:		This is 16x16 block which has its own subtype, which corresponds to other block types (range 3-9). Blocks should be decoded in the same way and then scaled twice.
	~~0 - pattern fill~~ (~~like block type~~ 3)
	2 - ~~"lossless" IDCT~~ )~~looks like scaled 8x8 IDCT though)~~
	~~3 - fill~~
	~~5 - masked fill (like block type 8)~~
	~~6 - raw coded 8x8 block~~ scaled twice

	=== 2 ===		=== 2 ===

Kostya: /* 1 */

2009-06-29T14:00:37Z

← Older revision		Revision as of 07:00, 29 June 2009
Line 88:		Line 88:

	=== 1 ===		=== 1 ===
	~~Divide~~ block ~~into subblocks and perform some operation on it~~:		Unlike other blocks, this seem to work with 16x16 block. It has a subcode for different real block types:
	0 - pattern fill		0 - pattern fill (like block type 3)
	2 - "lossless" ~~4x4~~ IDCT		2 - "lossless" IDCT )looks like scaled 8x8 IDCT though)
	3 - fill		3 - fill
	5 - masked fill		5 - masked fill (like block type 8)
			6 - raw coded 8x8 block scaled twice

	=== 2 ===		=== 2 ===