摘要 |
An image processing device, and a method and a program thereof are provided to enhance precision in detecting an object or a person from an image by considering appearance probability of the object or the person. An extractor extracts a feature quantity of an area probably including a recognition target in the image. A parameter storing part stores a parameter for the recognition target. A dynamic context storing part(42) stores context for the recognition target. A determiner determines whether the image of the area is the recognition target based on an operation result using the feature quantity, the parameter, and the context. If the area having a higher score calculated by using the feature quantity and the parameter than a threshold value is found, the determiner performs an operation for the recognition target corresponding to the image excluding the area by using the feature quantity, the parameter, and the context.
|