摘要 |
<p><P>PROBLEM TO BE SOLVED: To highly accurately restore a lost region of a motion picture signal which is compression-coded by inter-motion compensation frame prediction. <P>SOLUTION: An error concealment processing device creates an error dispersion holding map for all pixels in a frame at each time, and continues to update the map by using Kalman filter algorithm. More specifically, an estimated error covariance matrix P<SB>ax</SB>,<SB>y</SB>(t) and a diagonal matrix Q<SB>u</SB>(t) are set as initial values (S31), and an estimated error covariance matrix P<SB>bx</SB>,<SB>y</SB>(t) is obtained by using a state shift result A<SB>t</SB>(X<SB>x+dx</SB>,<SB>y+dy</SB>(t-1)) (S32). Then, K<SB>x</SB>,<SB>y</SB>(t) is obtained by using a diagonal matrix Q<SB>v</SB>(t) and an observation matrix H<SB>x</SB>,<SB>y</SB>(t) (S33), a motion vector X<SP>-</SP><SB>x</SB>,<SB>y</SB>(t) is obtained by using the state shift result again (S34), and an estimated present image X<SP>^</SP><SB>x</SB>,<SB>y</SB>(t) (S35). After that, the estimated error covariance matrix P<SB>ax</SB>,<SB>y</SB>(t) is updated by using the estimated original image (S36), and the processing is repeatedly executed from S32. <P>COPYRIGHT: (C)2011,JPO&INPIT</p> |