An approach to night-time HDR bracketing script

Sapog_topchet

An approach to night-time HDR bracketing script

« on: 17 / October / 2008, 23:25:20 »

Advertisements

I like to shot night city panoramas. HDR, sure.

There are some common things:
To make good pictures we shall minimize aberrations, so aperture shall be as tight as possible. (and keep IS off, sure)
Focus point may be at infinity.
Then, we shall take several shots at each exposure (3 shall be enought in most cases) to reduce ISO noise when merging image stack (or by Irani-Peleg rendering with ALE).
Then, shots shall be taken as fast, as possible to avoid ghosting. So, no 'noise removal' and dark frame subtraction.
And there should be a delay between commmand to start shooting and actual start: camera must be still. I found, even on a tripod 10 sec. may be not enough.

Q: Step of bracketing. What sane value may be set? 1EV? 5/3EV?

And night city has some special features.
First, there are illuminations. Lamps can easily beat out of average exposure to -4 or -6 EV (and sometimes even more). Dark areas which need +6EV are not so common, so there is assymetry. To make things simple, we can put the lower border near EV15 or EV16 (1/2000 or 1/4000 at f/4), regardless of measured luminance.
Second, there is some traffic on the roads. Red stop-lights traces never look good crossing facade of historical building. Such things shall be avoided. If possible, there shall be a kind of motion autodetection, but not to capture when motion detected, but to pause capturing and wait several seconds.

Getting all together:

Aperture number shall be the largest of possible
Focus shall be at infinity (and there may be saved time)
Noise removal shall be off
Number of shots at each exposure shall be tunable, with default 3, which shall be sane choise for most cases.
There shall be a delay before start, 15 seconds may be sufficient.
Shots shall start at the very high EV regardless of scene.
Then exposure shall increase by still UNKNOWN step
Upper bound shall be tunable, but +4EV completely enough for an average scene (with such a camera)
If possible, there shall be a way to pause shooting.

Script is intended to run on 570IS, and in future on 970IS.

Suggestions and criticism are welcomed. Grammar corrections too

Maybe somebody already made something similar?

What shall i take for the sample?

« Last Edit: 18 / October / 2008, 02:51:57 by Sapog_topchet »

Logged

fbonomi

469
A570IS SD1100/Ixus80

Re: An approach to night-time HDR bracketing script

« Reply #1 on: 18 / October / 2008, 02:24:43 »

Hi, this is a subject that interests me a lot.

I have being playing more with low-light situations, though (i.e. out of a town) so the requirements are abit different. See some discussion here: Night-time HDR with image stacking
The process is described here: Francesco Bonomi - Night-time HDR with image stacking

I love scripting, but in this case my opinion is: a script is just something that automates a process on your camera.

As long as you don't have the process very clear in your mind, setting things manually (with the help of CHDK) can sometimes help you as it gives you more flexibility.

I would suggest to do some tests, settings those parameters in CHDK and seeing the results.

As for the non-motion detect part, I am pretty sure it was discussed somewhere on the forum, but can't find the thread.... Also, a non-motion detect for long exposure would be rather useless, as it woud start shooting at time X, when there are no cars, but there is a high probability that at time X+5 seconds a car comes in the field.

« Last Edit: 02 / March / 2009, 06:25:32 by fbonomi »

Logged

Sapog_topchet

Re: An approach to night-time HDR bracketing script

« Reply #2 on: 18 / October / 2008, 03:38:54 »

Hi, Francesco.

Quote

I have being playing more with low-light situations, though (i.e. out of a town) so the requirements are abit different. See some discussion here: Night-time HDR with image stacking
The process is described here: Francesco Bonomi - Night-time HDR with image stacking

Yes, I've already read the whole " Creative Uses of CHDK " and I was quite impressed by your work

Quote

I love scripting, but in this case my opinion is: a script is just something that automates a process on your camera.

As long as you don't have the process very clear in your mind, setting things manually (with the help of CHDK) can sometimes help you as it gives you more flexibility.

I completely agree.
I like HDR, and i know some about digital imaging, but i have neither time nor desire to study it deep. I'm sorry that i can't make HDRi from image stack with a pencil and sheet of paper: I poorly understand computation of response curves, Paul Debevec's and Mark Robertson's equations, pixel weightening and so on. I use them, I've read some papers on them, but nothing more. And the same i can say on digital photography at all

Quote

I would suggest to do some tests, settings those parameters in CHDK and seeing the results.

Nobody made representive tests before? It hurts...

Quote

As for the non-motion detect part, I am pretty sure it was discussed somewhere on the forum, but can't find the thread.... Also, a non-motion detect for long exposure would be rather useless, as it woud start shooting at time X, when there are no cars, but there is a high probability that at time X+5 seconds a car comes in the field.

Yes. I thought of remote control which makes it able to interrupt capturing and make new shot when intruders are gone.

Logged

PS

157
A610 1.00f

Re: An approach to night-time HDR bracketing script

« Reply #3 on: 18 / October / 2008, 14:50:35 »

Logged

Sapog_topchet

Re: An approach to night-time HDR bracketing script

« Reply #4 on: 20 / October / 2008, 09:31:29 »

« Last Edit: 20 / October / 2008, 11:22:27 by Sapog_topchet »

Logged

PS

157
A610 1.00f

Re: An approach to night-time HDR bracketing script

« Reply #5 on: 21 / October / 2008, 18:35:21 »

Logged

Sapog_topchet

Re: An approach to night-time HDR bracketing script

« Reply #6 on: 22 / October / 2008, 07:26:42 »

Logged

dexotaku

Re: An approach to night-time HDR bracketing script

« Reply #7 on: 23 / October / 2008, 08:53:43 »

I also like to shoot night panos, and I'm literally just starting to explore the scripting abilities of CHDK [having just installed it].

I'm curious as to whether others reading this prefer open or closed aperture when doing night shots; I tend to leave whatever camera/lens I'm using's aperture as wide open as possible because stopping down the lens usually leaves aperture-shaped lens flare around all in-frame light sources. Leaving the lens wide open may mean ghosting in certain conditions but I prefer the soft flare it creates to the progressively more pronounced [usually hexoganal] flare [and accompanying pointillation/star patterns] induced when stopping down. I also prefer to keep my exposures as short as possible to minimise other artefacts such as motion blur.

This does mean more chromatic abberation [and depending on the lens this may interfere with point-matching for stitching with eg. Hugin] in each shot but I also try to maintain enough overlap between frames that the centre of each image dominates the final rendered pano.

I'm curious as to others' opinions on the matter.

Logged

dsvilko

Re: An approach to night-time HDR bracketing script

« Reply #8 on: 23 / October / 2008, 09:20:16 »

Quote

The nothing is cheap and widespread, but it's hard to get something out if it.
IP rendering does nothing with the nothing. It brings the way to merge several pictures in one.
To my experience with Canon compact cameras, 250 shots made in 'special iso3200 mode' on 970IS can be merged in single picture, which would be completelly noise/artefact-free and also sharper then the picture shot at iso80. Even more, it can look better then IP-rendered sequence of 3-5 iso80 shots.
It took about a week on my old 2.4GHz athlon x2 with 2GB of RAM and 8GB of swap. I'm sure, with better CPU and more RAM results could be better - i was forced to use lowest quality settings.

Correct me if I am wrong (I haven't read the actual paper, only the abstract) but Irani-Peleg rendering only works on the exposures that are shifted one relative to another. The exact shift needs to either be a priori known or precisely determined through image sub-pixel registration. You can't get a better resolution by combining multiple unshifted exposured (as in taken with a tripod). What you get is a reduction in noise/compression artefacts that can give an illusion of better resolution. You can use the good ol' ImageMagick to quickly average the exposures for the same effect. From my experience, when pushing the resolution to the limit, the biggest obstacle is actually the atmospheric turbulence that introduces small, random and local image shifts throughout the each image. These shifts then average to a slight blur with the actual resolution that is less than each stacked image (reduction in image noise is still worth it).

Logged