摘要 |
<P>PROBLEM TO BE SOLVED: To solve the problems of prior arts that it is difficult to automatically determine vanishing points for various images and that it is not easy to automatically configure a correct depth structural model to realize a stereoscopic vision without a sense of incongruity even when structural estimation in the law of perspective is applicable. <P>SOLUTION: In order to avoid a sense of incongruity from being felt from any non-stereoscopic image and also to determine a scene structure closest to reality to the utmost extent, three kinds of basic depth models are used, which indicate depth values for respective three kinds of basic scene structures. A composition ratio 8 is determined according to a high frequency component evaluation value of a received non-stereoscopic image from an upper image high frequency component evaluation section 2 and a lower image high frequency component evaluation section 3, and the three kinds of the basic depth models are composed in accordance with the composition ratio 8. An adder 10 superimposes the composited basic depth model to the R signal 9 of the non-stereoscopic image to generate depth estimate signal. The depth signal is produced by using the R signal of the received non-stereoscopic image and the composite basic depth model in one preceding non-stereoscopic image. <P>COPYRIGHT: (C)2006,JPO&NCIPI |