Can stacking be useful to hide things that move?

[arith]

Scenario; you are in a busy city and want to capture the lights, but not car headlights or thousands of blurry people. You want a high definition noiseless image of a city night scene.

Long exposure is out because of headlights; so I thought what would happen if I took a lot of short exposure images and stacked them in an averaging way.

So I did a google search and cannot find if this has been done or even if it will work. The reasoning behind it is if something moves then it may look a little like noise to a stacking program like PSE10.

But how long the interval and how many images? Cheers

[Tringa]

I may be misunderstanding something but I can't see how stacking a number of short exposure shots would help with car headlights.

Poorly illuminated moving people may not be recorded by a short exposure at night, but car headlights, lighted buses etc would be bright enough to show up, I think.

But as I said I may be misunderstanding something.

Dave

[Dave Humphries]

Hi Steve,

Dave's ideas are valid, I can't see this working "automatically" on headlamps, but by manual intervention, mainly cloning out short headlamp tracks on a wide angle shot, something might be possible.

The noise should average out OK though, there's a tutorial on that here.

Cheers,

[FrankMi]

Hi Steve, I think you can get the result you are looking for, but more likely with a photo merge type of stacking (rather than averaging). It can still be very tricky.

Long exposures can help when the reflected light is relatively weak compared to the exposure time, such as in people walking about. Automobile headlights will almost always be recorded no matter how long the exposure is due to the intensity of light.

I have had much better success with night shots in the city when taking many shorter exposures around 10-20 seconds and trying to time the shots so as to get a dead space (no automobile lights) in every part of the image. I can then merge all of the 'dead space' into one image and hopefully see only the empty streets.

The more the shooting angle is straight into the traffic, the less likely you will be to capture dead space so diagonal and cross traffic works better. You can also use stop lights to your advantage as it creates an ebb and flow to the traffic that tends to open up dead space between the vehicles. Lastly, you have a much better chance of capturing the dead space when the night traffic is at its minimum, such as around 4 AM.

You can shoot a large number of images and still not be able to get dead space in every location in the image sometimes because you can't remember unless you take notes, what parts of the scene you've been able to capture without any automobile lights.

In the image below, taken at dusk, there were automobiles fairly continously on all the city streets, but by timing the shots for the dead space I was able to eliminate most of the headlights and tail lights in the scene. Because I didn't get enough images of the bridge on the right hand side, you can still see one set of headlights peeping out from just under the bridge.



Another thought is to take images at different times of the night to get building lights, sky colour, etc. the way you want them but you don't want to leave your camera set up for hours on end.

One technique you can try is to setup your tripod and get the scene you want to capture repeatedly. Use the bubble level that comes with most tripods, pick a spot in the crosshairs of your sensor to aim the camera at, record the settings (particularly the focal length if using a zoom). Attach a plumb-bob to the tripod, then make a pencil or chalk mark on the sidewalk just under the bob so that you can get your tripod and camera in exactly the same position and settings so that you can leave and return to get additional images over time.

Hope this helps!

[arith]

Cheers Frank and both Dave's; I've read about averaging and even done it once. What I was thinking was something like in this experiment I just did.

The only thing I could think of that moves is the clock and the only flashing light is on my mobile

So I took 40 images on a 2 second interval using a cheap old card that only holds 43 pics but I don't want to use up the space on my UDMA card.

The first is the base image, 2nd is a two stack at 0.5 opacity and finally I got to 14 and think the second hand is pretty much gone.

It will be different for cars and people though; I want to photograph Fleet St toward the city

[DanK]

more likely with a photo merge type of stacking (rather than averaging).

Frank,

I'm note sure what you mean by this. Are you referring to a specific blend mode?

Thanks

Dan

[arith]

I have found out that you cannot rely on other peoples images to get an idea of the land; because I've found buildings in the wronge places.

I saw this torpedo style building in a photo with St Paul's looking from Fleet St, and I cannot see the GPO tower at the end of Euston looking towards Holborn.

Where is this torpedo shaped building?

http://www.bing.com/images/search?q=...0F7480C6918E4E

[Andrew1]

http://blog.lynda.com/2011/05/31/dek...-image-stacks/

If you don't have the particular package the presenter has perhaps you can figure out a work-around for the editor you have.

[James G]

Steve, the building is also referred to as "The Gherkin" and is located at 30 St Mary Axe (formerly the Swiss Re Building).It is a skyscraper in London's financial district, the City of Lond. It was completed in December 2003 and opened at the end of May 2004,and stands on the site of the former Baltic Exchange, which was extensively damaged in 1992 by the explosion of a bomb placed by the Provisional IRA.

I worked for BT before retiring and used to have meetings regularly at the BT headquarters nearby. Despite the number of times I went down to London, I never had the time to go in and take a look around, must remedy that soon!
:-) James

[arith]

Cheers Andrew, didn't see that. Thanks James, I saw the Gherkin large and green overpowering the dome of St Paul's in an image I found but looking at streetview it simply isn't there. It is a pity because I thought it looked impressive.

The George Inn in Borough High st is right next to 'The Shard', but this is a really busy and expensive place and the Shard is too high.

Definitely dificult to work out the best place to go to practice this kind of shot.

[FrankMi]

Frank,

I'm note sure what you mean by this. Are you referring to a specific blend mode?

Thanks

Dan

Hi Dan, it is the same technique that is used in manually doing focus stacking (as opposed to using Photoshop's Edit/Auto Blend Layers) except the pieces of the image are not selected based on how sharp that part of the image is but rather on the lack of vehicles and/or their lights.

[DanK]

Frank,

Thanks very much. I've used that for focus stacking, although usually I use Zerene. However, I am still puzzled by the process. Normally, the software selects from each layer or image the pixels that are in focus and rejects the rest. In this case, is it dropping the lights because they are moving and therefore not sharply focused? Or are you manually masking them?

Dan

[FrankMi]

http://blog.lynda.com/2011/05/31/dek...-image-stacks/

If you don't have the particular package the presenter has perhaps you can figure out a work-around for the editor you have.

Thank you for posting the link to Deke’s Techniques. I wasn't aware of the Smart Objects, Stack Mode, Median so I have learned something new. You've made that a good day for me!

Deke's process looks to be easier than a simple image merge for more complex objects but as Deke McClelland points out in the video, if there is an overlap with the random objects, then additional steps may need to be done to address this. In the case of Steve's automobiles and people, that would mean that no two images could have random objects that would overlap in the final composite result.

I will certainly give that technique a play as it is an option that I wasn't previously aware of. Thank you!

[FrankMi]

Frank,

Thanks very much. I've used that for focus stacking, although usually I use Zerene. However, I am still puzzled by the process. Normally, the software selects from each layer or image the pixels that are in focus and rejects the rest. In this case, is it dropping the lights because they are moving and therefore not sharply focused? Or are you manually masking them?

Dan

Yep! You can do focus stacking manually and I often do for touching up after Photoshop's Auto Blend Layers as it doesn't do a super great job. I've played with Zerene when I had their trial version but didn't have time to complete the process. If had switched to Zerene when I first got the trial version I might not have had to learn how to do the focus stacking merge manually! LOL!

It is the typical blending of two layers but in the case of city streets, I'm trying to blend together as much of the dead space as I can find rather than as much of the sharp parts of the image as I can find. It's the same process either way.

[FrankMi]

What I was thinking was something like in this experiment I just did.

The only thing I could think of that moves is the clock and the only flashing light is on my mobile

So I took 40 images on a 2 second interval using a cheap old card that only holds 43 pics but I don't want to use up the space on my UDMA card.

The first is the base image, 2nd is a two stack at 0.5 opacity and finally I got to 14 and think the second hand is pretty much gone.

It will be different for cars and people though; I want to photograph Fleet St toward the city

Hi Steve, I've played with that approach when I was experimenting with using that stacking technique to reduce noise in an image.

It will be interesting to see if that technique can be used to remove random objects like cars and people. I fear that like the Linda.com link that Andrew provided, it will be difficult where multiple images have cars and people in overlapping locations. I've not found an automated way to address that possibility as yet so I am hoping that your efforts can shed some light on a better approach.

[arith]

I think you are right Frank; looking at the experiment it looks like you need the area clear for about 15 frames, so say if you have a 6 second interval, 1.5 minutes before the next intrusion.

Very unlikely in this place, even at 4am Sunday morning.