An image processing method receives a sequence of image frames and generates a computer generated object. The method combines the object with the sequence of frames to generate a sequence of augmented reality images. The method divides each received image frame into a respective array of image motion cells, detects inter-image motion in successive image frames for each motion cell, and generates a motion object comprising one or more image motion cells. The image motion cells in the motion object correspond to a set of image motion cells detected as comprising inter-image motion over a threshold number of image frames. The method detects relative distance between the object and the motion object within the augmented reality images, and generates a point of interest within a current image frame so the object may appear to interact with an image region corresponding to an image motion cell at the point of interest.