Debanding of M240 at High ISO (DeNoise 5)?

[Jffielde]

Hello,


I have used Topaz DeNoise 5 for debanding of my Leica M240 to considerable success, but I would like to know if anyone has a good process worth sharing.



For those use use Topaz Denoise, what are your settings, particularly the "banding width"?


Any better programs? Thanks.

 

[Jffielde]

Thank you for your reply. The M240 has "push" settings (anything above 3200) that is acknowledged even by Leica to result in horizontal banding, particularly in the shadows. Yes, the images are all DNG and generally ISO4000-ISO6400. Virtually all "push" settings in the Leica M240 show at least some of this banding, so I thought that it might be common enough that others have figured out the optimal formula to minimize it. I just avoid anything above ISO 3200, but I would love to push a stop higher at times, if dealing with the banding was fairly easy. Topaz does a pretty good job, but I would still generally discard the images due to the banding. Thanks for your reply. Joe.

 

[Ronald M]

Mask the worst banding and treat it separately. Blur the mask edge with refine edge.

 

[borge]

The banding that the M240 produces can be easily fixed/removed in post by using Nik Dfine2 noise-reduction software / plug-in for Photoshop/Lightroom.

Just enable the "Debanding" feature in the Manual/More menu, and set a percentage of 30-35% removal and select vertical or horizontal depending on if you made the picture in portrait or landscape. Disable the NR itself: The grain is great at ISO 3200, the banding not so much. This easily eliminates the banding.

I almost never use Dfine with the automatic settings. I find them to be too agressive.

I rarely use the noise reduction part of Dfine. I only use the debanding tool for the M240, as I find the grain at 3200 ISO (the max ISO I use on my M240) completely acceptable.

The debanding feature in Dfine is a little bit hidden.

To find it, you have to open an image in Dfine and then:

1. Select the "Reduce" button/tab in the upper right of the interface. The default selection is "Measure" when you open Dfine.
2. Click the "> More" button on the lower right of the panel. This button is really small!
3. Here you enable "Debanding" and select "Horizontal" or "Vertical" and the percentage. You also get other options here, as Edge Preservation and JPEG Artifact Reduction, but I haven't used those.

For my use, I then go to the Contrast and Color noise sliders and drag them to 0% as I don't want any noise reduction, just debanding.

The Debanding feature is great, and works superb on images with banding. But it is hidden in the interface, which is sad, cause it's a great feature.

If you can't find the debanding option here, you need to update your Nik Software suite by downloading and installing it again.

 

[Jffielde]

Thank you.

 

[jaapv]

Just want to add: In general the banding is a lot worse because people tend to underexpose in low light situations, if only because exposure meters are easily fooled by specular highlights.
Finetuning your histogram -which is best done with the camera on manual, I find exposure compensation hit-or-miss- will often result in a much cleaner image.

 

[borge]

Just want to add: In general the banding is a lot worse because people tend to underexpose in low light situations, if only because exposure meters are easily fooled by specular highlights.
Finetuning your histogram -which is best done with the camera on manual, I find exposure compensation hit-or-miss- will often result in a much cleaner image.

It totally depends on the scene you are photographing in low light though.
When I'm walking around at evening or night time, photographing outdoors, I want my images to look like the scene actually looked, as in with low and dim light. Dark skies and dark parts of buildings, architecture, streets, rooms, etc. are usually always a part of a frame during night time photography. I wouldn't call making pictures that show the ambient light levels underexposure. But this is where the banding appears very very easily, even noticeable at 2500 ISO, easily noticeable at 3200 ISO and horrible at higher ISO's.

 

[jaapv]

It is better IMO to expose generously and get a healthy histogram, and pull the scene down in postprocessing to the degree of darkness you want than to deprive the camera of photons.
So yes, I would respectfully call your technique underexposure and it is certainly the root cause of the banding you experience.
Just to humour me - try exposing more liberally and pulling down in postprocessing a few times.

 

[willie_901]

While artifacts and noise can be filtered during post-production, nothing trumps recording data with the highest-possible analog signal-to-noise ratio.

The banding only depends on the datas' signal-to-noise ratio.

More signal (exposure when the shutter is open) improves the datas' analog signal-to-noise ratio.

Banding is an odd blend of artifacts (non-random uncertainties in the data) and electronic noise (psuedo-random uncertainties in the data). Both involve uncertainties, so for simplicity we'll just call the banding as noise.

As exposure increases, at some point the signal amplitudes and photon (a.k.a. shot) noise increase to where the artifact levels fall below the least significant bit of the analog-to-digital converter. Otherwise the banding artifacts are digitized.

For all digital cameras Prof. Emil Martinec (Physics, University of Chicago), described how to maximize exposure.

"(1) maintaining needed DoF, which limits how much you can open up the aperture;
(2) freezing motion, which limits the exposure time;
(3) retaining highlight detail, by not clipping wanted highlight areas in any channel.

Note that ISO is not part of exposure. Exposure has only to do with aperture and shutter speed. Maximizing exposure guarantees that one captures as many photons as possible subject to photographic constraints, and therefore optimizes S/N."

Items (3) is important to minimize the impact of banding and other artifacts. All highlights unnecessary for the photograph one envisions should be overexposed by the sensor. (This is not the same as clipping the signals in the ADC due to excessive signal amplification (a.k.a. ISO) although the results could appear similar.)

For night photography this means letting street lights, car headlights, etc, blow out... unless the details of those lights are important to the photograph. At some point the light levels are so low, significant underexposure is unavoidable – especially when a tripod is impractical.

All one can do is use the lowest possible ISO and bracket exposures (usually by 1/3 or 1/2 stops). A minimum ISO maximizes the SNR and bracketing provides options for highlight retention.

 

[jaapv]

No - it is not a part of exposure sensu stricto, as it does not influence the number of photons hitting the sensor and thus not the output of the sensor.
The ISO value is obtained by the amplification (simply put, there is a simple gain and a digital component here) of the output of the sensor.
Basically ISO on a digital camera is an electronic simulation of the ISO value of film, which is defined as the density of the film generated by a certain amount of light, something quite different.
One may argue that the ISO setting will in its turn influence the setting of the exposure parameters, but this is not the case here, as the object of the excersise is to increase the level of registered light on the sensor to improve the S/N ratio.

 

[willie_901]

Yes, ISO affects post-acquisition brightness.

Exposure affects brightness during data recording.

This is a profound difference when signal to noise ratio and dynamic range are important. This is often the case for night photography.

There is nothing one can do to improve the data after it's recorded. Data can be mathematically filtered (a.k.a. noise reduction) but noise filtering always degrades the data. The degradation is often minimal, sometimes it's not.

The same is true for dynamic range except noise filtering is irrelevant. Increasing ISO always reduces the analog dynamic range because the exposure (SNR) is reduced. Nothing can increase the DR of a single exposure after the data is recorded. However skillful post-production can squeeze the most out a raw file's information content.

Increasing ISO has serves three purposes for photographers.

1. One can use faster shutter times or smaller apertures because the light meter will estimate when the analog to digital converter will neither clip (too much signal) nor be underutilized (not enough signal level for optimum digitization). The light meter is not optimizing exposure (except at the sensor's base ISO). The light meter is optimizing brightness.

2. One can review focus and composition in-camera with a brightness (not exposure) that estimates the brightness one would see in a print.

3. Some cameras from certain brands produce a higher SNR above their sensor's base ISO. This has nothing to do with exposure. Instead the analog-to-digital converter technology adds electronic noise when the signal levels from the sensor happen to fall below a certain threshold. Increasing the signals' levels via ISO amplification enables those ADCs to operate with more effectively.

At any rate ISO has nothing to do with exposure whatsoever. I will send you a PM with links to reliable references from competing Forums and other sources if you'd like.

 

[jaapv]

They do; they are wrong. Your ISO button on your digital camera is just the volume control. Nothing more. It will allow you to have the same brightness whilst underexposing the sensor by the aperture and shutterspeed at the cost of dynamic range and S/N ratio (and ultimately detail and colour information).

 

[jaapv]

It is not possible to remove all misconceptions from the Internet...Some are stuck in the film concept. With film higher ISO will produce a high density on the negative with less light. On a sensor not, so ISO value is simulated by gain. Read the ISO norm. Film sensitivity is defined by density at a certain light intensity. Digital ISO value is defined as equivalency to film.
I suggest reading Willie 901’s post carefully. He has taken the trouble to explain things carefully.
However, to give Nikon their due, what they are explaining is the relationship between exposure and ISO to obtain the “correct" output level of the camera. The problem is that they simplified somewhat which confuses the issue.

 

[MCTuomey]

Bill, I think the point is that only 2 elements - aperture and shutter speed - can control the amount of light that strikes the sensor. The sensor's "sensitivity" is described and varied by ISO changes. An ISO change will influence aperture and/or shutter speed, but not the actual light quantity that one captures.

 

[jaapv]

No it is not. Exposure is the amount of light you admit onto your sensor. ISO is the amplification of the output of the sensor. And the amount of noise is determined by the amount of light that strikes the sensor. When you increase the ISO you are amplifying the sensor output, thus you are amplifying the noise. Just think of an old tape recording. If your recording input was too low you would have to increase the volume of your amplifier to get the same level out of your speakers. Result: the hiss of the tape became prominent.

 

[willie_901]

... An ISO change will influence aperture and/or shutter speed, but not the actual light quantity that one captures.

Yes. And the influence can be positive (maximum exposure) or negative (unintended or gratuitous low exposure) for the situation at hand.

Increasing ISO only minimizes the inconvenience or consequences of some shutter times, aperture choices.

The benefits of increasing ISO are:

Reviewing the results in-camera immediately with a useful brightness.

Analog-to-digital converter designs that require post-acqusition DC signal amplification to operate at their highest possible signal-to-noise ratio. In other words - minimize those ADCs' contributions to the total electronic noise in the raw data

Cameras with ISO-invariant (a.k.a. ISO-less) data-stream designs only benefit from the first of these.