Next, a specific process executed by the image processing apparatus (camera) will be described.
In the person detection process of the present disclosure, since personal detection based on image analysis can be performed even by a camera with a relatively low data processing function as well as an information processing apparatus with a high data processing function such as a PC, a significant reduction in computation which is necessary in the person detection process is implemented.
For example, in a case in which an attempt to perform the person detection from the captured image of the camera is made, in a general process of a related art, a feature quantity is calculated for each pixel constituting image, and a collation process of the calculated feature quantity and a learning dictionary in which feature data of an image of a person is registered is performed.
The feature quantity calculation process and the dictionary collation process are generally performed for all pixels in an image or all pixels in which the moving body is detected and are factors of increasing a processing load.
Further, persons of different sizes are photographed in a captured image in accordance with a distance from the camera. In order to detect image region of persons of various sizes, it is necessary to convert an image to images of different resolutions and carry out a collation process with a dictionary hierarchically for each of images of different resolutions.