Timelapse project - Creative Uses of CHDK - CHDK Forum supplierdeeply

Timelapse project

  • 6 Replies
  • 3918 Views
Timelapse project
« on: 09 / March / 2009, 16:50:41 »
Advertisements
A local bar/tavern brings in a local artist and a live model about once a month, and the artist sets up his easel and paints a portrait of the minimally-clothed model while everybody watches.  This is usually a three hour project with breaks along the way.  I thought it would be interesting to make a timelapse video of just the canvas, so you could see the creation take place before your eyes in fast motion.  Then I thought it might be cool to have the artist add narration to the video describing what he's doing at each stage.  So far, the bar owner is up for it, and I'm hoping the artist will agree.

So, armed with my mighty A590IS, I'm gonna give it a try.  Here are my lists:

Things to bring:

Camera w/4GB card
Tripod
White paper for manual white balance
Three sets of freshly charged Eneloop batteries
Script - Ultimate Intervalometer

Settings:

5 second interval
1600 x 1200 Superfine JPEG
Manual focus
Manual white balance
Manual mode -  fixed ISO, shutter speed, aperture
IS off
AF assist / red eye beam off
Review off
Mute on
Display off

Did I forget anything, or get anything wrong?  It looks like I can change batteries without moving the camera since the tripod mount is on the other side from the battery compartment.  However, I think for about $7 I can get the right power connector from Radio Shack and make up an externnal battery pack.  That would be safer.

I show 1600x1200 above, which is 2 MP.  But another alternative would be to just use 640x480, which should make an acceptable Divx video.  It also occurs to me that if I can persuade my tripod to do it, shooting in portrait mode might add to the picture information for the canvas, but I'm not sure what difference it would make after resizing and adding black borders to get back to a 4:3 video.

Anyway, I would appreciate any suggestions on this project.  This will be my first major timelapse. and actually my first CHDK project of any kind.  And let me also take this opportunity to thank all those who have contributed to CHDK.


Re: Timelapse project
« Reply #1 on: 12 / March / 2009, 05:45:38 »
If you can change the batteries without knocking the camera even slightly, I'd be very impressed! An external battery pack is a good idea and likely to be very useful with a camera using 2 AAs.

Programs such as Virtualdub can open the image sequence and downsize it efficiently, so I'd stick with the higher resolution. After all, you have a 4GB card!

I don't have my camera (an A640) with me at the moment so I can't find the exact settings, but you can force digital zoom on with optics at their widest focal length to give 'digital crop' - ie no up/downscaling or loss of detail, it's just as if you'd cropped a section of the full image. If I remember correctly:
Menu (Out of CHDK mode) -> Digital zoom - > fixed to 1.4
A '1.4x' icon appears at the top of the screen in blue - this means the image is being upscaled/downscaled
If I reduce the size of the image to M1 the '1.4x' icon turns white - this means the image is just being cropped.

The purpose of this? Simply to allow you to move the camera further away from the canvas without having to use optical zoom.

Sounds like an interesting shoot, have fun!

Re: Timelapse project
« Reply #2 on: 27 / March / 2009, 12:07:10 »
Staylo, I somehow missed your reply when you posted it.  Thanks very much for your comments.  I had completely forgotten about the "Digital Tele-Converter" function, but it works the same on my camera except I have the choice of 1.6x or 2.0x.

But I'm still not clear what's happening with it.  Is it just using a center segment of the sensor as the raw input, and then using that to produce various jpeg sizes?  It seems it would still have to upscale or downscale, unless the segment being used was exactly the same as the jpeg size.

I also have the blue/white changeover as I change the jpeg size, but the manual implies that this only means there's no "loss of quality" when white, not that it isn't scaling.  Actually, I think it must mean that when blue there is upscaling, but in white there is either downscaling or no scaling.

It seems I should be able to calculate whether 1.6s or 2.0x coincides exactly with one of the M sizes.

Anyway, thanks again for bringing this up.  If I understand, if you are shooting in one of the smaller sizes anyway, this lets you avoid the reduced effective aperture that would result from using the optical zoom, without losing any image qualilty.  And I guess that depends in turn on the idea that downscaling to jpeg from a moderately larger sensor segment size produces no worse results than downsizing to the same jepg from the full sensor input.  I'll bet that's a math question.  :-)


*

Offline fudgey

  • *****
  • 1705
  • a570is
Re: Timelapse project
« Reply #3 on: 28 / March / 2009, 08:21:16 »
When Digital Tele-Converter (let's call it DTC) is OFF, and you select a JPEG output resolution lower than maximum resolution (except full resolution wide angle, it's only cropped frop top/bottom), the camera must scale down. You can verify this by measuring a drop in burst shooting speed if your SD card is fast enough.

When DTC is enabled, the camera crops the sensor image, rejecting the sides. Thus it's not a digital zoom, it's just an area selector. The camera still outputs a JPEG with the selected resolution, so with your 1.6x or 2.0x DTC settings, the camera will scale up if you still try to use the maximum JPEG resolution. The color of the T symbol shows you when you have managed to select a JPEG output resolution that doesn't require scaling, causing no loss in image quality due to scaling (up or down) and the fastest possible burst speed for that resolution (if your camera is diffent in this respect, it's very easy to tell...).

As you've noticed, those 1.6x and 2.0x have been selected because they match your sensor size and JPEG reslolutions well.


Re: Timelapse project
« Reply #4 on: 28 / March / 2009, 13:35:00 »
Thanks, fudgey, but I still don't quite understand what you're saying.  With DTC enabled at, say, 1.6x, the A590 shows the T symbol in blue for L and  M1 (8mp and 5mp), but in white for M2, M3 and S (3, 2, and 0.6mp).   So it would seem that the only resolution which does not need to be scaled up or down is M2 - it's the largest size that's white, and the math at 1.6x works exactly right for that size.  But wouldn't M3 and S still need to be downscaled?  The same area of the sensor ends up in the picture regardless of jpeg resolution, so it would seem that only one jpeg resolution could exactly match the sensor window being used, and all others would have to be scaled one way or the other.

Well, in any case, it appears that at 1.6x DTC, M2 doesn't need to be scaled, and at 2x DTC, M3 doesn't need it either.  So it seems at least those combinations might prove to produce pictures as good as, or perhaps better than, using equivalent optical zoom at those same jpeg resolutions.  Unless - unless somehow the additional information provided by using the full sensor would produce a better picture after down-scaling than just capturing at the smaller size directly, and it's hard to see how that could be the case.

But if I'm right about the smaller sizes still needing to be downscaled, then there would still be a theoretical question whether using 1.6x optical zoom and downscaling from 8mp to, say, 2mp would be better than using 1.6x DTC but still downscaling from 3mp to 2mp.  I guess that's a question for sampling math wizzards, which I definitely am not.  My guess, though, is that it doesn't make much difference.


*

Offline fudgey

  • *****
  • 1705
  • a570is
Re: Timelapse project
« Reply #5 on: 28 / March / 2009, 14:30:21 »
I can't check from my camera right now, maybe I've forgotten how it works and it too indicates upscaling like yours. But anyway you're right. Btw, digic III propcase 92 (see http://chdk.wikia.com/wiki/PropertyCase) indicates the source image width... reading it with a script or debug osd propcase viewer and comparing with your JPEG resolution setting is another way to make sure you're not scaling.

So it seems at least those combinations might prove to produce pictures as good as, or perhaps better than, using equivalent optical zoom at those same jpeg resolutions.  Unless - unless somehow the additional information provided by using the full sensor would produce a better picture after down-scaling than just capturing at the smaller size directly, and it's hard to see how that could be the case.

Downscaling should produce a better image because the camera (if it's a 10Mpixel one) doesn't actually have 10 million RGB pixels... it has 5 million green pixels 2,5 million red pixels and 2,5 million blue pixels in a Bayer matrix. This means that full resolution JPEGs actually can't hold that much detail for all colors even if the optical parts were up to it...all pixels are just cunningly interpolated from adjacent single color pixels. What this means is that when you downscale a zoomed in image to a low resolution instead of cropping a part from zoomed out image, you should have more information to work with unless it's destroyed by blurry optics or noise (zooming in reduces light, which is very important since our cameras are highly noisy...).

Re: Timelapse project
« Reply #6 on: 30 / March / 2009, 13:18:38 »
Well as soon as the snow melts, I'm going to set up a test and see if I can tell the difference between:

A.  2x optical zoom, no DTC, 2mp jpeg

B.  2x DTC, full wide optical zoom, 2mp jpeg

In theory the second shouldn't need downscaling.



 

Related Topics