Adaptive Temporal Kalman Filter For Real Time Video Filtering

1
Adaptive Temporal Kalman Filter For Real Time
Video Filtering

• Members
• Raju Muchanthula
• Sasikanth Munagala
• Michael Yaw Appiah

2
Video Filtering

• Optimal Approach
• adaptive 3-D spatial temporal filter
• Complexity
• Proposed Solution for real time implementation
• Independent processing in temporal domain

3
Assumptions

• The process model
• s(k1)a.s(k)w(k)
• a is constant
• w(k) is process noise, xero mean Gaussian random
  process
• x(k)s(k)v(k)
• v(k) is additive zero mean Gaussian white noise

4
Kalman filter
5
(No Transcript)
6
(No Transcript)
7
(No Transcript)
8
(No Transcript)
9
Defining the Parameters
10
Temporal Kalman Filter
11
(No Transcript)
12
Results
Threshold(G2) Values for Motion Detection
Confidence Level () G2 G
99.9 10.827 3.29
99 6.635 2.577
98 5.412 2.325
95 3.841 1.96
90 2.814 1.646
13
Simulink Demo
14
Conclusions

• Independent processing of image-sequences in
  spatial temporal domains
• Simple effective in perfomance
• Using Kalman filter algorithm with pre-spatially
  non-linear filtered frames.
• Adaptive filter easily implemented in HW

15
The End

• Thank You

16
(No Transcript)