| 摘要 |
A device and method for reducing the resolution of an input image G showing a three-dimensional scene which is captured by a monitoring camera 3, characterised in that the resolution reduction is dependent on the object distance in the 3D scene and/or the perspective distortion of the 3D scene. The object distance is the distance between the camera and the scene objects in the 3D scene. Areas of the image with a large object distance have their resolution reduced less than the areas of the image with a small object distance so that distant objects and close objects that are the same size have a similar resolution for subsequent object segmentation, object following or motion detection 10. The resolution reduction may be adjusted by analysis of camera parameters, by modelling a scene reference model and/or a datum plane or by combining with other pre-rectifications of the input image. The distorted and resolution-reduced input image may be used for image analysis where object-describing metadata is formed. This resolution reduction method enables real time video processing because it reduces the processor load by undersampling in the resolution uniformly in the horizontal and vertical directions before image processing. |
