/966 LMS adaptive filter Weight vector Input sample d(n)=W'(n-1U(n) vector e(n)=d(n)-w'(n-1U(n) d(n) u(n) an） V(e2)= F-block =-2el m7 wm=wn-)-we2o》 W(n-1) Weight Error update D feedback e(n) loop A W(n) loop WUD-block 2021年2月 11
2021年2月 11 LMS adaptive filter F-block D WUD-block u(n) d(n) e(n) + - Error feedback loop Weight update loop ˆ d n( ) W(n-1) W(n) ( ) ( 1) ( ) ˆ d n W n U n T e(n) d(n) W (n 1)U(n) T ( ( )) 2 1 ( ) ( 1) 2 W n W n T e n W eU W e e W T T( ) 2 2 2 Weight vector Input sample vector
1.2.5 motion estimation /96 Motion estimation is used in interframe predictive coding. N-1N-1 s(m,n)=∑∑x(,)-y(i+m,j+n) i=0j=0 u=min (m.n s(m,n) Motion (0,0) vector Best N-1,N-1) matched block N+2p 2021年2月 12
2021年2月 12 1.2.5 motion estimation Motion estimation is used in interframe predictive coding. Motion vector Best matched block (0,0) (N-1,N-1) N+2p N+2p 1 0 1 0 ( , ) ( , ) ( , ) N i N j s m n x i j y i m j n u min (m,n) s(m,n)