摘要 |
<p>PROBLEM TO BE SOLVED: To obtain an encoding method suited to omniazimuth picture multi- view encoding by generating a predictive picture without seam on the assumption that the end of a referring picture is continued to the other end, and encoding a difference between a generated omniazimuth picture and an inputted omniazimuth picture. SOLUTION: A feature extracting part 101 extracts a vertical edge from an omniazimuth picture. A feature tracking part 102 obtains the corresponding relation of a vertical edge. A six-direction extracting part 103 convert from an edge to an observing angle. Memories 104 to 106 store the angles of a reference point A and moving points B, C. A camera motion calculation part 107 finds a parameter by repeating arithmetic from observation. Thus, a camera motion parameter is estimated by assuming a parallel line with a lens in a center on the projecting surface of a normal camera by using the camera and tracking feature movement on the line. Namely, at the predict-encoding of the omniazimuth picture, motion compensation combining a right end and a left end is executed to encode the picture efficiently.</p> |