Unless I'm totally mistaken...

The image is acquired from right to left, in vertical bars. As this happens, the wheels are turning. The spoke that was at 3:00 when the front of the wheel was being imaged is moving around clockwise toward 4:00, 5:00, 6:00 as the image continues to be recorded from right to left.

Meanwhile, I assume that the *camera* is moving or panning from left to right, to follow the riders; this would blur the background and keep the riders in the frame.

...and, for more info:

The 12:00 spoke is moving "against" the image acquisition direction, so is imaged very briefly and very sharply.

The 6:00 spoke is moving "with" the image acquisition, so it remains in the frame for a couple of vertical imaging line samples. This makes it look thicker and fuzzier.

A beautiful photo and an intelligent question! Adding to gregm123456's good points: The effect is known as “Roget's Palisade Illusion”. A detailed explanation is given here:

In the present case, the shutter timing works like a fence.

no you've both got it wrong. Back then spokes were that shape under load. They were made of birch. It was thought that the bend in the spoke would result in extra forward movement, known as teh "gfree lunch" lacing technique, it lasted until someone pointed out that it was a load of old cobblers and that use of exotic materials (titanium and carbon fiber) was the only thing guaranteed to make a cyclist go faster. That and drugs.

/runs

Similar effect have been observed using a high speed camera to film archery.

See this thread on archery-interchange.

http://www.archery-interchange.net/forum/compound-bow-discussion-q/18242-look-what-happens-release.html

At the time it caused at lot of discussion because it suggested that the bow was not set-up correctly. As the other posters have suggested, it's linked to the shutter or in the archery case above to the CCD in the camera. It's an artefact of the camera capture.

Most modern finish line cameras use a video sensor only one to three (b/w vs color) pixels wide. This vertical 'slit' is precisely aimed at a fixed point on the road, and as the riders pass through that 'slit', scans of that line are recorded at a very high rate of speed and assembled into a conventional image.

The spokes are not recorded all at once, but as they cross the line.

The nipple end of the spokes at the front of the wheel are caught exactly where they should be, but as that spoke continues through the slit, the rest of the spoke has advanced - making the spoke appear curved.

Same goes for the back side of the wheel. The hub end is caught first by the camera, and the rest of the spoke will be moving clockwise as it is recorded - hence, the upward bend to the nipple.

The suggestion of this being a "Palisade Illusion" is correct. The example used on this page is illuminating if you set the Palisade option to "move" - and imagine the wheel passing through just one vertical slit.

Images of wheels taken in the golden age of photography began showing this effect when the first horizontal tracking focal plane shutters came into use. A fast shutter speed has the first and second curtain of the shutter crossing the film close together making a vertical slit. Depending on the direction of the wheels travel (left or right), the direction of the shutter slit (up or down, left or right) and the stillness of the camera (not panning) the curved spoke illusion would be oriented differently and the wheels would not be round, but would 'lean' forward or backward.

The additional effect of the image being stretched or compressed lengthwise is a due to the rate the scans are recorded compared to speed the riders are moving. There is a rider speed and scan rate setting in the camera software that compensates for this.

A side note: the camera must be aimed at a lighter area at or near the finish line (never the black part) so the background will be light enough to see the absolute front edge of a dark colored tire against the background. This means that most video finish line images you see were NOT aimed at the actual line, but before or after it. That is, unless your actual finish line is reversed from the norm - a white line on a black background.

gregm123456 :

For those who doubt you, the motion of the spokes describes a cycloid. See here:

http://mathworld.wolfram.com/Cycloid.html. The forward velocity of a spoke approaches zero as it nears 6:00, and so is imaged several times by the camera's right-left scan.

8hangtime: I think it's unlikely that the camera used a CCD linear array, as I don't believe it would be mechanically possible to step the array rapidly enough to avoid blurring the slower-moving components of the picture, such as the rider's feet at the top of the pedal stroke, Software could correct the foreshortening the riders and bicycles, but I don't think it would compensate for any blurring of the feet, just as it didn't compensate for the blurring of the spokes. Furthermore, you'd expect the feet to be foreshortened more than the rest of the bicycle, since they are moving twice as fast as the bike at the top of the stroke, and it's unlikely that any compensation algorithm would treat the feet as a special case.

I think this is more likely a 2D CCD, which would be sequentially sampled as a raster - in this case in vertical lines right to left. It's not sampled quickly enough for the spokes, but it's plenty fast enough for the rest of the bike & rider.

As for the 'leaning' wheels, so familiar from cartoons as an indication of speed, this can be traced back to an image of a race-car taken by French photographer J.H, Lartigue during the ACF Grand Prix in 1912. Lartigue's camera used a relatively slow vertical focal-plane shutter, not the horizontal shutter you describe, which is used in more recent film cameras. Lartigue's vertical shutter produced a horizontal slit moving bottom-to-top. Lartigue panned the camera to follow the car, but the wheels were still 'bent' forward, while onlookers in the background appear 'bent' the other way.

The photograph can be seen here:

http://www.artnet.com/PDB/PublicLotDetails.aspx?lot_id=426226497&page=1

To me these finish images look like they are taken by a static Linescan camera as described by 8hangtime. This would be why the background appears as horizontal lines - its a 1 pixel wide image of the finish line repeated side by side across the generated image. The riders would be imaged as they cross this line.

The speed of the riders as they cross the line will effect how stretched they appear in the image. As will the acquisition rate of the camera and whether square pixels are used to display the image. Presumably the camera is set to acquire lines at a rate that is appropriate to the expected speed of the finishing riders, so that the riders and bikes appear with close to natural proportions.

datafusion - why would the feet be moving twice as fast as the bike at the top of the stroke? You are thinking of the top of the wheel. The feet would only be moving a small proportion faster than the bike due to gearing. This is what gears are for after all! IF you were usign a 1 to 1 gear ration AND the cranks were the same diameter as the wheels you woudl be correct (or if you had a large rear sprocket and small chainring you could achieve what you suggest with normal length cranks). - So with normal gears the feet at the top position area only moving a small percentage faster than the bike and rider so the shortening would not be that obvious.

If you think through a wheel crossing the static linescan imaging line - the first post by gregm123456 pretty much explains it - you are imaging a spoke as it rotates through different angles, at the same time the point of interception between the spoke and the imaging line moves along the length of the spoke. The 3:00 spoke appears level at the nipple end, as the wheel moves across the imaging line the spoke rotates down to 6:00 whilst you are imaging along the spoke up to the hub. So between the front edge of the wheel and the hub crossing the imaging line, the spokes image goes down then back up. I disagree with the first post regarding the 'panning' assumption. To me it is clear that the camera is not panning it is static and aligned with the finish-line.

Caught this with my iPhone a few weeks back: http://twitpic.com/46yrh

Taken by me while riding at around 28km/h.

GiantPete - Sorry, quite right about the pedal rotation. I was hung up on the cycloid thing. A 175mm crank has an angular velocity of about 1m/s at 110rpm. Given that these guys are probably doing on the order of 13 m/s, the velocity increase at the top of the stroke is fairly negligible - so the foreshortening is only on the order of a single US shoe-size (although his feet look big enough as it is :-) ).

Didn't realize I could click on the picture for a bigger image. And, yes, I now think you and 8hangtime are right about the stationary slit. Note the positions of the left and right feet of rider 523: it appears to me as if the cranks are not quite 180 degrees apart, which is consistent with the left foot being imaged significantly later than the right.

The big question is, of course, what kind of wheels is 523 riding? Here's my vote:

http://www.ritcheylogic.com/dyn_prodfamily.php?k=135225

-Rick

Datafusion beat me to it. It's the shutter plane that goes across the field of view and records as it goes. If something moves in the shot at that time, it will record the movement as it goes so distorting the object. I also like Mr Lartigue too with his bendy car wheels.

Whilst some present day or ex photography students are in the house; is Tim De Waele (spellling??) the best Cycle Photographer? I think he has a great eye.

Cheers Jerry

Actually, I'm very familiar with how this photo was created. I personally secured it after the race from Rich Weiss of PhotoFinishPlus. It is neither a single shot nor a frame from a high speed film strip of the finish but an image comprised of many narrow frames shot at 4,000 frames per second, far beyond the speed of the racers. The software assembles the individual, narrow frames into images like you see there. It is amazing that the high speed distorts the fast-rotating spokes in that manner but not the slower-moving racers, but with that kind of speed and image-capturing, it is not surprising.