| 摘要 |
A method for characterizing objects within an image is presented. According to an embodiment of the present invention, a representative color of an object is first identified from an image. The contour of this same object is then determined by identifying regions of the image where the identified color changes to another color. Color and contour data from other images of the same object may be obtained using the above described method to compile a range of colors and a range contours which represent the object. This color and contour data (known as the "object signature") is then associated with a name for this particular object and stored. Once an object signature has been defined for a particular object, a database of images may be searched to identify images which include objects matching the object signature. For example, once an object signature for a mountain is determined, this object signature may be used to search a database for all images which include a mountain.
|
