FAST-moving balls can appear as blurry streaks in photographs and video stills. That is a problem for commentators examining disputed line calls and coaches studying how well golfers and table-tennis players control balls. Now a group of scientists in Italy have worked out how to determine a ball鈥檚 path and spin from a single blurry image.
Motion-blurred images contain far more information about a ball鈥檚 trajectory than frozen ones, say and his colleagues from the Polytechnic of Milan, who have developed a way to extract this data.
Looking at the blurred streak of a moving ball in a photograph, it鈥檚 easy enough for software to detect the angle at which the ball is moving left or right and up or down in relation to the camera. The difficulty comes in working out how the ball is moving towards or away from the camera. This can be done by measuring the changing width of the blur, since the ball will appear smaller when it is further away. Existing software cannot do this because a motion-blurred image has transparent edges, confusing edge-detection algorithms.
Advertisement
Giusti and his colleagues have developed a new algorithm based on the idea that a blurred image is equivalent to a series of sharp images added together. They calculated what a series of brief exposures would look like and were able to work out a formula that describes the transparency of the blur towards its edges.
The new algorithm uses this formula to determine where a ball鈥檚 edge is and then to calculate the change in its distance from the camera (). Exploiting information such as the colour of the ball and its background, the software can compensate for variations in lighting, which may affect how transparent the ball appears.
Knowing the exposure time and the size of the ball, the team can work out the speed and direction of a ball from relatively short smears. If the ball has some surface pattern, their software can even determine how it was spinning. This capability could be a useful training aid for sports such as golf in which players use the spin of the ball to control its trajectory. It should also be cheaper than existing devices, because it uses only one stills camera while other systems need coordinated video cameras to follow a ball鈥檚 motion. 鈥淭his would be great. 3D is very expensive,鈥� says Chris Swanner of in California, which develops video training systems.
Giusti says that a commercial application such as judging disputed line calls may still be some way off, as the team has yet to test its method on real sports in the field.
