H.264 / MPEG-4 Part 10

1
H.264 / MPEG-4 Part 10

• Nimrod Peleg
• March 2003

2
Encoder
3
Encoder
4
Encoder
5
Encoder
6
Encoder
7
Previous Standards
8
H.264 Brief review

• Goal
• Develop a high-performance video coding standard.
• Start from zero.
• No backward compatibility.
• Assumptions
• Block based.
• Software implementation.
• Network friendly.

9
H.264 Encoder
10
Transform

• DCT like.
• Integer arithmetic only.
• without multiplications.
• Optional use of a 4x4 transform block size.

11
Quantization

• Compounding quantization step.
• Thirty-two different quantization step sizes.
• The step sizes are increased at a compounding
  rate of approximately 12.5.
• Different quantizer for luminance, chrominance.
• Two different coefficient-scanning patterns.
• The simple zigzag scan.
• The double scan.

12
Frames Store

• The H.264 standard offers the option of having
  multiple reference frames in inter picture
  coding.
• Up to five different reference frames could be
  selected.
• Resulting in better subjective video quality and
  more efficient coding of the video frame under
  consideration.
• Using multiple reference frames might help making
  the H.264 bit-stream error resilient.

13
Motion Estimation Compensation

• Motion Estimation is where H.264 makes most of
  its gains in coding efficiency.
• Quarter pixel accurate motion compensation.
• Translation only.
• The standard does not determine which algorithm
  should be used.

14
Motion Estimation Compensation

• Seven optional modes.

15
Entropy Coding

• H.264 has adopted two approaches for entropy
  coding
• Universal Variable Length Codes (UVLCs)
• One table.
• Context-Based Adaptive Binary Arithmetic Coding
  (CABAC)

16
Intra prediction

• 9 optional prediction modes for each 4x4 luma
  block.
• 4 optional modes for a 16x16 luma block.
• 4 optional prediction modes for a 8x8 chroma
  component.
• One mode for 4x4 chroma block.

17
Summery

• Transform
• IDCT
• Optional use of a 4x4 transform block size.
• Quantizer
• step sizes are increased at a compounding rate of
  approximately 12.5.
• Two coefficient-scanning patterns.
• Motion estimation and compensation
• Translation only.
• A number of different block sizes are used for
  motion prediction.
• Quarter pixel positions are used for motion
  prediction.

18
Summery

• Frames store
• Multiple reference frames may be used for
  prediction.
• Entropy coding
• Only one regular VLC is used for symbol coding.