• Please use real names.

    Greetings to all who have registered to OPF and those guests taking a look around. Please use real names. Registrations with fictitious names will not be processed. REAL NAMES ONLY will be processed

    Firstname Lastname

    Register

    We are a courteous and supportive community. No need to hide behind an alia. If you have a genuine need for privacy/secrecy then let me know!
  • Welcome to the new site. Here's a thread about the update where you can post your feedback, ask questions or spot those nasty bugs!

Alternative Process: HDR tonemapped Video

Hi Folks,

Interesting experiment: http://vimeo.com/14821961

The postprocesing is a bit crude, and some of the end result is not how I would like to see it, but the potential is there.

It's nice to know that the CMOS technology already allows to extract images with multiple "ISO" setiings from a single capture, although it means moving more data at the same time, so that might require more processing power to combine with high FPS settings.

In-camera high scene contrast image improvement is already available in some (P/S) still cameras (e.g. iContrast for Canon, active D-Lighting for Nikon ), although I don't know if it is anything more than tone adjustment. For video it's only a matter of time ...

Cheers,
Bart
 

Doug Kerr

Well-known member
Hi, Bart,

It's nice to know that the CMOS technology already allows to extract images with multiple "ISO" settings from a single capture . . .

I had never thought of that before. Important point!

If we digitized the analog output from each sensel to a higher bit depth, then that could be exploited in one stroke!

I have to ponder the meaning of that!

Maybe after breakfast.

Best regards,

Doug
 

Mike Shimwell

New member
Hi, Bart,



I had never thought of that before. Important point!

If we digitized the analog output from each sensel to a higher bit depth, then that could be exploited in one stroke!

I have to ponder the meaning of that!

Maybe after breakfast.

Best regards,

Doug


I think the key here is that higher isos are achieved in CMOS by amplifying the analogue output voltage before AD conversion. This gives lower noise than increasing AD bit depth at present. For example, the 1Ds3 has a 14 bit ADC, so (given a well capacity of just over 60,000 electrons or approximately 2^16) it reaches unity gain at iso 400 (i.e. 1 electron equals one increment to the raw data). However, for low light shooting iso 800 is quieter than underexposing iso400 by one stop and pushing in post. Therefore the amplifier must contribute less noise than other elements of AD conversion at least to that level. In practice too, a 1Ds3 doesn't deliver 14 stops of dynamic range as the total noise is too high even at iso 100.

So, if we could build a 16 bit ADC with low enough noise to sample each well accurately to the full depth then that would give 16 stops of DR in one stroke. In effect all the information there is in the well is captured.

Given that no one has delivered such an ADC to the consumer market so far, we give up information by converting at a single iso, but may recover some of this by reading out several times at different isos.

I have used a related approach when shooting HDR occasionally. If shooting moving water it is helpful to use a consistent shutter speed (and aperture of course), so making equal exposures at 100 and 800 iso can help extract the full range from from the sensor. Of course, if you need more range then you need to vary actual exposure as well. This can be refined further if you are shooting a typical waterfall in a gully situation by adding a further lower exposure for the sky using 100 iso and a shorter shutter speed. This can then just be blended in to the really bright bits that (usually) don't move as quickly as the water.

Best

Mike
 

Doug Kerr

Well-known member
Hi, Mike,

I think the key here is that higher isos are achieved in CMOS by amplifying the analogue output voltage before AD conversion.
The same is true of CCD sensors; the difference there is that we don't (ordinarily) have the option of contemplating each sensel voltage in sequence several times with different amplifier gains.

But we could certainly have a different kind of amplifier-ADC module that would in effect do that in time-parallel (perhaps as a way to get better "bit depth" while keeping amplifier noise at bay).

Therefore the amplifier must contribute less noise than other elements of AD conversion at least to that level. In practice too, a 1Ds3 doesn't deliver 14 stops of dynamic range as the total noise is too high even at iso 100.
Indeed, I skipped over the matter of how amplifier noise (wherever measured) varies with amplifier gain.

Thanks for the nice discussion.

Best regards,

Doug
 
Top