BTIC2D - cr88192/bgbtech_misc GitHub Wiki

BTIC2D was an experimental DCT based codec.

Each macroblock consists of an 8x8 DCT Y block followed by 4x4 DCT UV blocks.

Coefficients are encoded in zigzag order, using AdSRice.

• Values greater than or equal to zero encode coefficient values.
  • These values have an additional sign-fold.
• Values less than zero encode a run of zeroes.

Possibility 2:

• AdSRice: Values!=0 encode coeff value.
• 0 escapes a zero run (AdRice)
  • 0: EOB
  • 1+: Encodes the number of zeroes.

Possibility 2-1:

• Store an image with 1 pixel per macroblock;
• Colorspace is YUVDyuv;
• DCT blocks encode pixels relative to the YUVDyuv space;
  • Y=Cy+Ly*Dy
  • U=Cu+Ly*Du+Lu
  • V=Cv+Ly*Dv+Lv
• Add multiple block codings.

Block-level image data uses prediction and linear quantization.

• Block-Level Cmds
• Cy, Dy, YBlk8x8
• Cu, Du, UBlk4x4
• Cv, Dv, VBlk4x4

 A B
 C x
 P=(3*B+3*C-2*A)/4

Exp Rice Table

  0- 7  0- 7  0
  8-11  8-15  1
 12-15 16-31  2
 16-19 32-63  3
 ...

Exponential coding with AdRice used for coefficients, with the value being sign folded. Zeroes remain special, and are used for EOB or escape-coding runs of zeroes.

Use a mask table.

• xvuy-VUYs
• s: sticky
• YUVyuv: indicates which components are updated.

Block Formats

• 0, ZigZag+DCT
• 1, ZigZag+WHT