Faking a Strip Camera

[Yoyo]

Given my suggestion of a strip camera in the thread about the linear guides, and the article I dug up and linked, I started thinking about how to fake a strip photo with a digital camera. Here’s my train of thought:

Use video mode of course to capture lots of individual frames in rapid succession. Either 720P at 60 FPS for fast subjects, or 1080P at 30 FPS for better detail or perhaps 24 or 25 depending on the particular situation. Split the movie into individual frames. Crop a column a pixel (or a few pixels) wide from each frame (the same column from each frame) and assemble them all into one image side by side. If the frame rate and column width in pixels are properly selected for the speed of the moving objects, then the subject should assemble very nicely, because even though not every column is looking at the same place, relatively it doesn’t matter because it’s all about the relationship between time and space, not either of them individually.

Well, I set up the tripod and shot a couple movies at 640x480x60fps, and all that’s left to do is take them apart and put them back together. I’ll let you know how it turns out.

[Yoyo]

Argh! Why must Canon create impossible to read Quicktime crap MOVs? And why does all the MOV to MPG conversion software suck? Can anyone recommend a program to convert a MOV to individual stills? And why on earth are there so many different encoding, compression, video, audio, and container formats that all seem to be incompatible with each other? Argh!

I have learned however that stuffed animals flying across the screen at 6″ per frame, i.e. 30 feet per second, are moving much too fast to make this work.

[fluffybunny]

This is why I quit doing video. Good luck, we’re all counting on you.

[Yoyo]

Ok, so getting a bit mathematical now.
The camera is set-up with a 4′ level across the frame so I can measure stuff. There are 31 inches in the frame, but I’ll round up to 32 to keep things simple. Shooting 640×480 gives a horizontal resolution of 640/32 = 20 pixels per inch (I know not stellar). Now, if I were to take a column of 1 pixel wide, i.e. 1/20″, from each frame, i.e. 1/60 sec., the subject of the image would need to be moving at 1/20″ / 1/60 sec. or 3 inches per second. Wow, that is some slow moving object. Now if I took a 2 pixel wide column, it would go twice as fast, etc. Perhaps if I used dental floss to tie the stuffed animal and drug it across the frame? Let’s give it another try!

[Yoyo]

Well, here it is. The final product. A strip photo taken with a standard dSLR. I was moving the animal at about 4 ips, so for the final I rescaled horizontally by 75% to correct the aspect ratio.

[Yoyo]

And here’s a still photo to show you the set-up and that the resolution of the strip photo isn’t so bad consider it was captures as MOV, converted to WMV, extracted to BMP, croped as BMP, stitched into left and right pano JPGs, those two stitched into one JPG, and then resized. The yellow streak above at the right is a tape measure I paper-clipped to the tail to pull the beaver. You can sort of see the paper-clip on his tail above. The strips were taken from about the 39″ mark on the level in this image, between the table and the door frame..

[orionid]

I’ve tried this. I still have more options to try with it. In the Nikon firmware, the lines are scanned vertically in each frame, so with a 50mm lens and a train moving at 55 mph, there was about 18 inches of slant between the top of the train and the bottom of the train for any given frame. My next attempt will include orienting the camera vertically to attempt to negate that effect. I am still planning on using strips of 10 or 15 pixels, though.

You’ve actually made more progress than I have on this, again, like with the other thread, I haven’t had time since this past summer.

[Yoyo]

If you’re having problems with rolling shutter, I would recommend skewing the image in post-processing instead of rolling (rotation around lens/fore-aft axis) the camera body. But now that I think about it, you’re exactly right orienting the camera in portrait would totally eliminate the rolling shutter effect as a single column of pixels would be captured all at the same time. It would also have the added benefit of greater resolution.

Also, in case anyone was wondering, I used Irfanview to separate the video into stills, then batch crop into columns, and finally stitch the columns into a single image. I could still use a recommendation for a .MOV frame grabber.

[Author]

Posts