Use your camera to create slit camera (slit-scan or timeline) images.

You get the horizontal distortion when the camera's framerate and the motion of the subject aren't exactly in sync.

- If the subject is moving faster than one pixel-width per frame-time-length, then it'll be horizontally compressed, as in the vehicles shown in the first post.
- If the subject is moving exactly one pixel-width per frame-time-length, then it'll not be compressed or stretched, as in the walking people in the first photo.
- If the subject is moving slower than one pixel-width per frame-time-length, then it'll be horizontally stretched.... as the subject lingers in the slit's field of view, the slit will capture the same part of the subject over and over, and they'll be stretched horizontally.

Regarding the direction of travel:
If the subject is moving away from the sun, the shadow will enter the slit's field of view before the subject. If the subject is moving toward the sun, the subject will enter the slit before the shadow. We can tell that we have some cars moving left-to-right, and some cars moving right-to-left, but without knowing the relative location of the sun, we can't identify which set is moving left-to-right, and which set is moving right-to-left.
... however, if the photos were taken in America, we know that the cars in the far lane were moving right-to-left, and the cars in the near lane were moving left-to-right.


Martin, thanks for posting, this is a great discussion. Give the ol' noggin a workout.

waterrockets wrote in post #15328799
I love looking at a bike throw in a slit-scan image. The winner often appears to be on a longer bike because he or she timed the throw to maximize as the front wheel hits the line, then they pull the bike back under them as it crosses the line, which slows the speed of the bike and gives it more columns of pixels for the final image. For the same reason, they look to have really long forearms because they move with the bike, but their torso and hips are going a more constant speed in spite of the throw, so they don't get longer.

... Isn't this backwards? Perhaps I'm misunderstanding the term "throw."

Assuming a constant camera framerate, If the front wheel moves through the slit faster than the back wheel, then the front wheel would appear horizontally compressed or the back wheel would appear horizontally stretched. It's moving so fast that some parts slip through the slit's field of view without being captured. Or, the back wheel is moving so slowly that some parts of the wheel are photographed twice - so they show up in two vertical lines in the final image - so it's stretched.

Compare it to the image in the first post. The cars are compressed, the people are not - because the cars are moving faster than the people and faster than the camera's framerate.

[edit] ah, I think I got it - the bike is actually moving more slowly as it crosses through the slit?
[edit 2] Maybe not. If it's rapidly decelerating, the back of the bike should appear longer than the front of the bike.

nathancarter wrote in post #15329277
[edit 2] Maybe not. If it's rapidly decelerating, the back of the bike should appear longer than the front of the bike.

I'm confused now

Looking at this photo (below), it does do as you suspect, where the rear wheel appears longer than the front wheel (rider #942 for 3rd place is the only one throwing the bike at the line, and the front wheel is oval with a more circular rear). There are a lot of photos with the winner's bike both in front of and behind the 2nd place rider, as 2nd place comes from behind and is crossing the line much faster.

I also love the way spokes look on slit-scan images.

nathancarter wrote in post #15329256
- If the subject is moving faster than one pixel-width per frame-time-length, then it'll be horizontally compressed, as in the vehicles shown in the first post.
- If the subject is moving exactly one pixel-width per frame-time-length, then it'll not be compressed or stretched, as in the walking people in the first photo.
- If the subject is moving slower than one pixel-width per frame-time-length, then it'll be horizontally stretched.... as the subject lingers in the slit's field of view, the slit will capture the same part of the subject over and over, and they'll be stretched horizontally.

For a non-panning camera, it should be a nonmoving object that won't be stretched or compressed. For any moving object, it depends on how it's moving relative to the slit (which is moving).

The objects with the reversed shadow are backward in the image; they have to be. The shadow is essentially part of the object; it cannot be treated differently. If the object overtakes the moving slit, this will happen (each vertical slice is read and then frozen in place, sort of like peeling off a sticker but in reverse).

Here's what I like: http://www.largeformat​photography.info …i-Lartigue-and-his-camera both the car shot and the dancing couple.

Anton, in this exercise, it's a non-moving vertical slit that captures objects and shadows that pass through its tall, skinny field of view. Not a vertically panning horizontal slit as shown in your link (though that's another very interesting exercise in itself).

In this style of slit camera, a non-moving object will be stretched all the way across the final image. The slit's field-of-view will capture the exact same vertical sliver every time, so when those columns of pixels are assembled into the final image, it will give an illusion that looks sort-of like a panning blur.

Look at the backgrounds in the images posted in this thread.

Also, look at the spokes in the bicycles posted above, compared to the spokes in the car in Lartigue's image.

Got it. That explains a lot that wasn't explained anywhere!

nathancarter wrote in post #15329766
Look at the backgrounds in the images posted in this thread.

Yeah, a properly placed finish line camera will have a white finish line right in line with the slit, making a completely white background.

Hi, apologies for intruding as I work mostly with video and dont own a C dig cam.

As I was playing with a similar goal I stumbled upon this forum/script.
Great work Martin !

What I did was use regular video footage and converted it to a higher frame rate ( 500 currently) and create similar picture with avisynth scripting. Now I am trying to figure out ways to indicate time differences and markings

converted video in your script :

doggyjumper wrote in post #15369868
converted video in your script

Very pleased you managed to run the script! Thanks for the feedback.

doggyjumper wrote in post #15369868
converted it to a higher frame rate

Not really sure how increasing frame rate of a video file would be any different from stretching the exported image in photoshop. I can't be sure from your examples. How "fast" was the car going?

doggyjumper wrote in post #15369868
Scan line locations are not at same position !

You can change scan line position in my script.

I had a wild idea that shooting in portrait orientation via several mirrors parallel to the raster scan could enable multiple slits to be taken from the same frame which would be focused on the same location but at different times - however looking at the blur of high speed objects in a single frame I feel this would be a lot of work for insignificant improvement.

Thanks again for sharing your results. Hope to see more

Hi Martin,

Thanks for sharing this program. I am interested in trying to work with still images as well as video. Please can you tell me how to configure the program to accept stills, I shall be using a slow-rotating turntable to gather the images.

David

Panphoto wrote in post #15923728
Hi Martin,

Thanks for sharing this program. I am interested in trying to work with still images as well as video. Please can you tell me how to configure the program to accept stills, I shall be using a slow-rotating turntable to gather the images.

David

Should be quite easy to do and could give higher resolution results, though I assume you would need quite a few images 800-4000? If I get a mo, I'll take a look.

Hi Martin,
Yes, I would expect the number of images to be up in the thousands. Perhaps it might be necessary to render the image in chunks of say, 1000-2000 at a time, and then stitch them together. I think high quality JPEG's might be the best solution. Here's an image I made using your program! I hope you like it. David

I don't think my picture posted correctly, would some kind soul 'hold my hand' through the proceedure?

Panphoto wrote in post #15938628
I don't think my picture posted correctly, would some kind soul 'hold my hand' through the proceedure?

I don't use PhotoBucket, so can't give specifics of how to post from there.

You can easily post from your computer -- you'll want to set your image size to the POTN limit (1024 pixels at the tallest in this case) and then set your jpeg Quality to give you a file of within 150KB. Some apps have a function to limit your file size (Lightroom and Photoshop do, for example) but with others you'd set the Quality to a "modest" level (somewhat below medium) then save and check the resulting file size.

Once you get a proper size, you can open the POTN Attach window -- open the Advanced Reply dialog and in the top toolbar you'll see a little paper clip icon around the center of the toolbar. Click it, then Browse to find your file, then click to Upload. When done, you can close the window.

To show your pic, when you are in the text of your reply and your cursor is placed where you want the pic to show, click the Attach icon again. A little window will pop up with your jpeg listed at the bottom. Click your file and an Image ID will show surrounded by the IMG tags, and the pic will show up when you post the reply.

Tony, Thanks for your help!