Video Compression - PowerPoint PPT Presentation

1 / 15
About This Presentation
Title:

Video Compression

Description:

Full screen = 640x480 required 7 million bit (about 7M bit) ... In order to prevent flicker, moving pictures often have to refresh at high rates ... – PowerPoint PPT presentation

Number of Views:102
Avg rating:3.0/5.0
Slides: 16
Provided by: mich114
Category:

less

Transcript and Presenter's Notes

Title: Video Compression


1
Video Compression
2
Introduction The need of compression
  • True color image 24bits per pixel
  • Full screen 640x480 required 7 million bit
    (about 7M bit)
  • Smooth video clip at least 25 frames/second
  • Therefore, 7 million bits x 25 175 million
    bits/second required (about 175 Mbps) !!!
  • Also, 1 min video is required about 10.5 G byte
    of data

3
Introduction
  • Video Compression deals with the compression of
    visual video data.
  • Video contains spatial and temporal redundancy.
  • Usually reduce the redundancy using lossy
    compression
  • Image compression techniques to reduce spatial
    redundancy
  • Motion compensation techniques to reduce temporal
    redundancy

4
Motion Picture Compression
  • Moving pictures are essentially a series of
    static images.
  • This being the case many of the techniques for
    static image compression will work for each of
    the individual frames.
  • Further compression can be achieved by taking
    into account the nature of moving pictures.

5
Moving Picture Compression
  • In order to prevent flicker, moving pictures
    often have to refresh at high rates (sometimes at
    least 55Hz).
  • The differences between subsequent frames in a
    series will not necessarily be great and so
    techniques for Motion Compensation in order to
    assist the compression.

6
Motion Compensation
  • Compare two sequential frames in a series, note
    the differences between them, and code only the
    residuals.
  • Each compressed frame is rebuilt by adding
    decoded residuals to the prediction based on the
    previous frame.

7
Motion Compensation Block Matching
  • Partition image into blocks
  • Previous frame is searched for each block of the
    current frame.
  • If a match is found, the location of the block
    (the position vector) in the previous frame is
    coded.
  • A trial predicted frame is built by using blocks
    from the previous frame, displaced by the
    movement vector.

8
  • Video sequence -gtPicture -gtSlice -gt Macroblock -gt
    Block with 8 pixels in both dimensions

9
Motion Compensation
Second Frame
Motion vector
Residual
First Frame
Prediction
10
Motion Compensation
  • The predicted frame is subtracted from the next
    actual frame, and the residual image is encoded
    for transmission.
  • At the receiving end the process is reversed, the
    predicted frame is built from the previous frame,
    and the residual image is decoded and added to
    the predicted frame.

11
Motion Compensation Process
  • First Frame normally is encoded without motion
    compensation
  • This frame is called Intra frame (i-frame)
  • The other frames are called P-frame and are
    predicted from the I-frame of P-frame that come
    immediate before it.
  • Therefore the output of the prediction schemes
    can be described as IPPPPPP

12
Motion Compensation Process
  • Frames can also be predicted from future frames.
  • It is called B-frame
  • The future frames have to be encoded before the
    predicted frame.
  • Therefore the encoding order does not necessary
    match the real frame order.
  • Such process is predicted from two directions
  • Therefore this coding scheme can be describe as
    IBBPBBPBBPBB

13
Motion Compensation
  • Searching an entire frame for a matching piece
    requires much processing power.
  • Searching is therefore restricted to a small
    region surrounding each new block, this is
    defined by an encoding parameter called the
    search range.

14
Motion Compensation
  • Transmitting long motion sequences by means of
    frame prediction has two problems
  • If an error occurs in one Frame, it will
    propagated through future frames, until a change
    obliterates it.
  • In many cases video must be editable.
  • Should be possible to cut any group of frames
    from a video sequence and use them in another
    sequence.
  • With predictive codeing a video sequence must
    begin at the first frame.
  • Both problems are solved by periodically sending
    a full frame that is not a prediction (eg. Every
    10 frames). These Frames are called I-Frames.

15
MPEG flavours
  • MPEG-1
  • Designed for data rates of up to 1.5 Mb/s
  • Single speed CDROM rate
  • T1 data communications.
  • MPEG 2
  • Designed for higher rates 15Mb/s and higher
    quality performance including HDTV
  • MPEG 4 and 7 under development.
Write a Comment
User Comments (0)
About PowerShow.com