摘要 |
<P>PROBLEM TO BE SOLVED: To solve the problem that it is difficult to make the decision of a dissipating point automatically to various images and, moreover, even if a perspective structure estimation is a compatible case, it is not easy to achieve a stereoscopic vision which constitutes a right depth structure model automatically and does not have a sense of incongruity. <P>SOLUTION: In order not to impress the sense of incongruity to any images and at the same time to make the decision of a nearer scene structure as actually as possible, a basic depth model of three kinds of quantifies is used for the depth values about each of the three kinds of the scene structures used as an elementary. A non-solid image which performs scaling to a predetermined image size is used, a synthetic ratio is determined to the high frequency component evaluation values of an input non-solid image from the high frequency component evaluation part of an upper part and the high frequency component evaluation part of a lower part, and according to the synthetic ratio, three kinds of basic depth models are compounded. In an adder, after the compound basic depth model and the R signal of the non-solid image are compounded, scaling is performed again and a final depth signal is formed. <P>COPYRIGHT: (C)2006,JPO&NCIPI |