intERLab

1
DVTS Video Frame Rate Adjustment Based on Motion
Detection
By Nawat Kamnoonwatana nawat_at_interlab.ait.ac.th
2
Introduction(1)

• DVTS in narrow bandwidth environment
• Lower frame rate
• Fixed frame rate for a single DV stream

Statistics Akimichi Ogawa, Katsushi Kobayashi,
Kazunori Sugiura, Osamu Nakamura, Jun Murai,
Design and Implementation of DV based video over
RTP, May 2000, Packet Video Workshop 2000
3
Introduction(2)

• How to improve the existing method?
• Bandwidth consumption can be improved by
  exploiting the motion in DV stream.
• Dynamically adjust frame rate based on amount
  motion is required.
• Two required elements
• Motion Detection Mechanism
• Frame Rate Adjustment Mechanism

4
Motion Detection(1)

• Simple mechanism
• Detect the amount of motion not the moving
  region.
• Keep computational time as small as possible.
• Temporal Difference of two frames
• Decode DV frames back to YUV color space
• Only the Y-component is used to calculate the
  temporal difference.

5
Motion Detection(2)
6
Motion Detection(3)

• In order to visualize the overall motion in the
  video stream the following factors were
  introduced
• Stepping factor Compute the difference only at
  certain period of frames.
• Moving average filters window size Apply moving
  average filter to smooth out the data.

7
Motion Detection(4)

• Stepping factor of 10 and Moving average window
  size of 10

8
Frame Rate Adjustment(1)

• Interpret the output of motion detection and
  adjust the frame rate accordingly.
• Requirements
• Work in any environment (i.e. lighting
  conditions)
• Do not consume CPU resources inappropriately
• Must have smallest effects on video quality

9
Frame Rate Adjustment(2)

• Threshold Method
• Threshold cannot be fixed since different
  environment gives different motion detection
  values.
• Defined from motion detection output range (i.e.
  maximum and minimum values)
• Sensitive to abrupt changes in lighting
  condition.

10
Frame Rate Adjustment(3)

• Threshold is defined by global maximum and
  global minimum.
• Frame rate is adjusted only when the motion
  detection output crosses
• threshold for a period of time (e.g. 1 second)

11
Frame Rate Adjustment(4)

• Confinement Method
• Confine the motion detection output within a
  certain range regardless of its lighting
  condition.
• Solution use the change in motion detection
  values.
• Interpret the values directly to the preferable
  frame rate using a simple scalar quantization.
• Eliminate the problem of sensitivity of the
  abrupt change when using Threshold method.

12
Frame Rate Adjustment(5)
13
Frame Rate Adjustment(6)

• Factors that control the output of Confinement
  Method
• Sensitivity
• How sensitive to the amount of motion to be
  detected.
• Defined by the range of confinement.
• Quantization
• Number of steps How many frame rates that
  reflect the amount of motion.
• Maximum minimum frame drop ratio
• Both must be adjustable by the user.

14
Frame Rate Adjustment(7)
15
Frame Rate Adjustment(8)
16
Future Work

• Fine tune parameters
• Which should be fixed and which should be
  adjustable by user.
• Combine parameters if possible
• E.g. Sensitivity Factors
• Optimize the process (i.e. required buffers and
  process time)
• E.g. Substitution for Moving Average filter