| 摘要 |
PROBLEM TO BE SOLVED: To adjust panorama image by correcting a wide-angle image, in particular, distortion of the image by one or a plurality of user. SOLUTION: In order to teach a neural network, a scale pattern defined by the array of scale points is arranged, and placed in the field of view of a wide-angle image device. Thus, the scale image is captured by the device. Each direction of the view point of the scale point and the position of the point in the scale image are used for the data to teach to neural network, and each direction of the view point is corrected according to the position in the image. Next, in order to generate sub image free from any distortion from arbitrary wide-angle image with distortion, the array of the pixel position of indication is expressed as a corresponding set of the direction of the view point, and converted to the set to the position corresponding thereto in the wide-angle image having the video characteristic at the position allocated to the pixel corresponding to the sub image by using the neural network. |
