| 摘要 |
<p>PURPOSE:To delete useless information generated due to noise in a background part and to improve the efficiency of coding by coding only the inside of a speaker's area obtained by segmentation. CONSTITUTION:It is supposed that a speaker moves as shown at time t0, t1, t2, movement compensation is executed between the images of time t1 and t2 to find out movement and a moving vector. It is supposed that the movement between the two images mainly depends upon a mouth part, its moving vector is shown by plural arrows in Fig.A. An independent arrow part existing on a background part is regarded as a moving vector generated by the noise of the background, a block generating the moving vector is set up as an effective block and a block generating no moving vector is regarded as an invalid block. Since the speaker's part is segmented and coded, useless coding information generated due to noise on the background or the like can be removed and the efficiency of coding can be improved.</p> |
