摘要 |
PROBLEM TO BE SOLVED: To specify an image of a zoomed-up/zoomed-down scene from among a plurality of time-sequentially consecutive images and to produce a zoomed image based on the specified image. SOLUTION: An image selection section selects consecutive two image data from dynamic image data. A feature point extraction section extracts a feature point of a reference image. A corresponding point search section searches a corresponding point that is corresponding to a feature point of a target image, based on the feature point of the reference image. A zoom power detection section detects zoom powers and a distribution of the zoom powers for each combination of the corresponding feature point and its corresponding point and if the distribution is less than or equal with a threshold, the selected two image data are specified as image data of the zoomed-up/zoomed-down scene. A zoomed image production section constructs pixels of the zoomed image from pixels of the reference image or the target image for pixels near an image center point based on the specified two image data or composes pixels of the reference image and pixels of the target image for pixels in a periphery of the image, constructs the pixels of the zoomed image from the synthetic pixels and generates data of the zoomed image. COPYRIGHT: (C)2005,JPO&NCIPI |