| 摘要 |
Disclosed are an object region extractor device, method, and program, which simultaneously are capable of improving non-extraction of object regions in three-dimensional spaces that arise in errors in extraction of object regions from images, and of minimizing mistaken extractions of object regions in three-dimensional spaces. The object region extractor device comprises an effectiveness assessment quantity acquisition means, which is for deriving an assessment quantity that increases (or decreases) as an angle, which is formed from two back projections from among back projections of one point of an object that is respectively photographed by each of m image acquisition means (where m is a plural number) and that are obtained from the images that include the one point and that intersect at a location in a three-dimensional space, approaches a right angle; and an object region determination means for determining that the location is a region that the object region occupies and outputting the assessment result, when the assessment quantity is greater than (or less than) a prescribed value. |
