Panasonic G9 in low light

BDR-529

Mu-43 Top Veteran
Joined
Jun 27, 2020
Messages
707
You have another option and an alternative to using either Topaz Denoise AI or DXO Deep Prime, which can work with or without a tripod. M43 and the G9 in particular make it relatively simple to take multiple images in a burst, and subsequently stack & average in Photoshop to dramatically reduce noise.
Of course you do need to use Photoshop for this. There may be other software that can do the same, but I haven't heard of it.

I have Topaz Denoise which I use for pretty extreme cases like ISO3200 and above but it has a tendency to create defects of it's own so I prefer Darktable profiled denoise if I can do with ISO800-1000

And you don't need Photoshop for averaging out noise from multiple shots. This is what some HDR algorithms are able to do. It might be that they have some clever build-in NR system .

(edit: there seems to be a huge difference between HRD implementations and some might even increase noise. I've had excellent results for example when scanning 35mm slides with MFT camera and macro lens using HRD. Not only does this method capture details which are not visible otherwise but result has less noise at least when I use Darktable.)

Try feeding 3-5 shots with the same exposure to HDR or better yet use small exposure bracketing like 1/3 step to take advantage of increased dynamic range while you are at it.

Totally free Darktable has excellent HRD but I can't imagine how this module could be made any more complicated and confusing to use for users who are new to HRD.
 
Last edited:

L0n3Gr3yW0lf

Dad Pun Joke Master Over Nine Thousand Meme Lord
Joined
Jul 31, 2013
Messages
1,119
Location
UK
Real Name
Ovidiu
Ovidiu, I cannot speak for the G9, but this shot is with my E-M1 MkI that has a Panasonic 16 MPx sensor.

View attachment 877304

The full size LSF JPEG that is completely untouched, OoC, is here. Be warned, it's a 9MB file ...

https://canopuscomputing.com.au/zen2/albums/Photography/HighISOdemoshots/E-M1_JAK_2016-_2112448.JPG

I reckon that it's not all that bad for an OoC JPEG. It would certainly print absolutely fine at A3 size, and almost certainly at A2 size. Even at 100% on my screen, it looks reasonable.
Very beautiful picture :)
From what I have seen over the years I believe, now, that most M4/3 sensors can do high ISO well under one condition: good or direct light (as you have noticed you cat was well light by the side lamp on the left side).
There are many moments where my E-M1 Mark III (which should be the 2nd best at high ISO as tech evolution wise, only better would be the 10 MP from the GH5s) struggles a lot at high ISO because the light is very defused and low (one such example would be thick underbrush on a cloudy day).
It could be a limitation if of not being able to demozaic accurately with limited light/information, which would introduce a lot of chroma noise. And it would make it hard for the JPEG engine to make good understanding of what's in contrast, details, and what can be noise reduced, background, where the image is turned into a mush.

A third issue that applies only for the E-M1s (and possibly more Olympus cameras in the future as they "upgrade" their stack to the E-M1 20Mp sensors) is hot pixels, I can get hot pixels to show uok in pictures as soon as 1/2 sec. Shutter speed in HHHR. It could be an issue with the dual cross PDAF that leaves less room for the heat from light gathering to dissapate/expand OR manufacturing difficulties to bring the sensor wafers at high enough quality
(Which makes me curious of something, maybe we could make a thread for all E-M1 III/II/X and share a dark frame to campare how much hot pixel the cameras have and to what extend it could be an issue)
 

John King

Mu-43 Hall of Famer
Joined
Apr 20, 2020
Messages
2,693
Location
Beaumaris, Melbourne, Australia
Real Name
John ...
Very beautiful picture :)

Thank you, Kind Sir :flowers_2:.

From what I have seen over the years I believe, now, that most M4/3 sensors can do high ISO well under one condition: good or direct light (as you have noticed you cat was well light by the side lamp on the left side).

There is a standard lamp behind the chair. However, look at the EXIF: f/6.0 at 1/20th and ISO 6400 is hardly good light. The bulbs are also "Warm white" (I will not have a daylight colour temperature light in or on the house. There are excellent medical reasons why lights at night should have a colour temperature less than 4,000ºK, but that's a long and complex subject, which I have informally studied for about 20 years).

Here is another, with no direct light.

E-M1_MkII_JAK_2020-_M226102_Ew.jpg
Subscribe to see EXIF info for this image (if available)


There are many moments where my E-M1 Mark III (which should be the 2nd best at high ISO as tech evolution wise, only better would be the 10 MP from the GH5s) struggles a lot at high ISO because the light is very defused and low (one such example would be thick underbrush on a cloudy day).
It could be a limitation if of not being able to demozaic accurately with limited light/information, which would introduce a lot of chroma noise. And it would make it hard for the JPEG engine to make good understanding of what's in contrast, details, and what can be noise reduced, background, where the image is turned into a mush.

Interesting reasoning, but I think that you are quite wrong about this. I have quite a few photos where the lighting is even, and lower exposures than this, where the results are more than just acceptable. One should always consider how to get as much exposure as possible = maximise TIME * APERTURE, while keeping the ISO as low as possible. This is the secret to minimising noise.

A third issue that applies only for the E-M1s (and possibly more Olympus cameras in the future as they "upgrade" their stack to the E-M1 20Mp sensors) is hot pixels, I can get hot pixels to show uok in pictures as soon as 1/2 sec.

Turn on in-camera NOISE REDUCTION (NOT Noise Filter, that's different). NR causes the camera to take a dark frame after exposure at long exposures. This is then subtracted from the image exposure frame to remove hot pixels.

Shutter speed in HHHR. It could be an issue with the dual cross PDAF that leaves less room for the heat from light gathering to dissapate/expand OR manufacturing difficulties to bring the sensor wafers at high enough quality

No.

(Which makes me curious of something, maybe we could make a thread for all E-M1 III/II/X and share a dark frame to campare how much hot pixel the cameras have and to what extend it could be an issue)

Just a word of warning - PIXEL MAPPING should never be carried out when the sensor is hot. Follow the manual. Make certain that you haven't used the camera for some hours. Then turn on. Wait the time recommended in the manual (I THINK that this is half a minute, but CHECK!). Then perform pixel mapping once. I find that my cameras need this doing about every 12-24 months.

I hope this clarifies a few things for you.

I plan the reprint the original image tomorrow. I ran the head cleaning program on my XP-970, and that cleared the blocked Light Cyan head. I don't know how it got blocked. I printed about 8 6x4" shots just two days ago, and they were absolutely perfect??
 

archaeopteryx

Gambian sidling bush
Joined
Feb 25, 2017
Messages
1,655
A few addons to what @John King said.

(which should be the 2nd best at high ISO as tech evolution wise, only better would be the 10 MP from the GH5s)
Things like this are commonly assumed but, as is also often the case, this particular assumption is not well supported by measurement.

It could be a limitation if of not being able to demosaic accurately with limited light/information, which would introduce a lot of chroma noise.
In a strict sense, variance in demosaicing would require the Bayer pattern to change as a function of ISO. It doesn't. To a first order approximation, I think what you're referring to is mainly the use of one red and one blue versus two green pixels per Bayer cell becoming more obvious at lower dynamic ranges. Additionally, quantum efficiencies aren't constant across colour channels, usually meaning DNR/SnR is lowest in blue.

There are second order effects around ISO variant and ISO invariant behaviors if you want to get into that. The main thing is firmware/software changes noise management at higher ISOs typically have larger effects than hardware ones within a given base ISO.

It could be an issue with the dual cross PDAF that leaves less room for the heat from light gathering to dissipate/expand
The base process you're looking for is photon to charge conversion. It's quantum mechanic rather than thermal.

And it would make it hard for the JPEG engine to make good understanding of what's in contrast, details, and what can be noise reduced, background, where the image is turned into a mush.
Not fundamentally, as there's nothing structurally unique about a jpeg engine. SOOC jpegs are the output of an in camera raw development and there's abundant evidence manufacturers vary development parameters as a function of ISO. Essentially, a lower signal to noise ratio at the pixel level requires more pixels be aggregated to obtain a given SnR in an image processing operation.

How much this matters depends on what the operation is but the basic thing is it makes it harder to produce raw developments which pixel peep well. That is distinct from pattern and structure recognition—AI understanding, though understanding is maybe a bit of a stretch here—of image content.
 
Last edited:

L0n3Gr3yW0lf

Dad Pun Joke Master Over Nine Thousand Meme Lord
Joined
Jul 31, 2013
Messages
1,119
Location
UK
Real Name
Ovidiu
A few addons to what @John King said.

Things like this are commonly assumed but, as is also often the case, this particular assumption is not well supported by measurement.

In a strict sense, variance in demosaicing would require the Bayer pattern to change as a function of ISO. It doesn't. To a first order approximation, I think what you're referring to is mainly the use of one red and one blue versus two green pixels per Bayer cell becoming more obvious at lower dynamic ranges. Additionally, quantum efficiencies aren't constant across colour channels, usually meaning DNR/SnR is lowest in blue.

There are second order effects around ISO variant and ISO invariant behaviors if you want to get into that. The main thing is firmware/software changes noise management at higher ISOs typically have larger effects than hardware ones within a given base ISO.

The base process you're looking for is photon to charge conversion. It's quantum mechanic rather than thermal.

Not fundamentally, as there's nothing structurally unique about a jpeg engine. SOOC jpegs are the output of an in camera raw development and there's abundant evidence manufacturers vary development parameters as a function of ISO. Essentially, a lower signal to noise ratio at the pixel level requires more pixels be aggregated to obtain a given SnR in an image processing operation.

How much this matters depends on what the operation is but the basic thing is it makes it harder to produce raw developments which pixel peep well. That is distinct from pattern and structure recognition—AI understanding, though understanding is maybe a bit of a stretch here—of image content.
I can admit I am wrong and I am more than happy to learn something new. It was an assumption and I will not hold on to :)
 

BDR-529

Mu-43 Top Veteran
Joined
Jun 27, 2020
Messages
707
From what I have seen over the years I believe, now, that most M4/3 sensors can do high ISO well under one condition: good or direct light (as you have noticed you cat was well light by the side lamp on the left side).

I would rephrase that because now it sounds a bit strange. Like why on earth would you use high ISO and suffer from all that noise in the first place if "good light" is available.

My version is "M4/3 sensors (which is practically that same sensor: IMX269, IMX270 and IMX272) can capture good quality images at high ISO if there is so much light that the scene is resonably bright to naked eye and high ISO was used just to get a short shutter speed"

AI denoise programs like Topaz can deliver spectacular results in cases where the subject was sort of brightly lit to the naked eye but not so much when the whole scene was truly dim and underexposed.

I know that in theory it should not matter whether sensor got so few photons due to short exposure or because there weren't many of those around to start with but these are just the real world results I get with MFT camera.

This is very common problem for anyone who is shooting sports indoors or late evening on a "brightly" lit outdoor venue. It might look bright to the naked eye but m4/3 camera will strongly disagree if you try to use anything faster than 1/100s.
 
Last edited:

John King

Mu-43 Hall of Famer
Joined
Apr 20, 2020
Messages
2,693
Location
Beaumaris, Melbourne, Australia
Real Name
John ...
Here is a photo of the two A3 prints. The one on the right is of Lizzie, taken with my E-510 at ISO 800. It was printed from the raw file.

M216547.JPG
Subscribe to see EXIF info for this image (if available)
 

saladin

Mu-43 Hall of Famer
Joined
May 29, 2015
Messages
2,254
Location
Melbourne
Real Name
jason
Taken last night in real world, twilight gloom. Iso5000 and a still too slow shutter speed, max aperture of f/6.3 was the biggest problem. As was a propensity to AF on the foliage when in "animal detect" mode. . So i switched to use pinpoint AF , it took some persuading but eventually locked on the duck rather than the reeds, but it does highlight that the G9 has a few tools you can use to find ways around various problems. It has some noise, and 1/40th is simply too slow for even minor tremors and movement in the subject, but i don't mind the mood. It may be suited to monochrome processing.

P1023941_cropped.jpg
Subscribe to see EXIF info for this image (if available)




Edit: Like this. Imported into Silver Efex, application of "antique plate" preset, a little structure adjustments. ISO 5000 isn't ideal, but its workable if you have a vision or end product in mind, i think.


P1023941_cropped_Nik_nik.jpg
Subscribe to see EXIF info for this image (if available)
 
Last edited:

Brownie

Mu-43 Hall of Famer
Joined
Sep 3, 2018
Messages
3,572
Location
SE Michigan
Real Name
Tim
Taken last night in real world, twilight gloom. Iso5000 and a still too slow shutter speed, max aperture of f/6.3 was the biggest problem. As was a propensity to AF on the foliage when in "animal detect" mode. . So i switched to use pinpoint AF , it took some persuading but eventually locked on the duck rather than the reeds, but it does highlight that the G9 has a few tools you can use to find ways around various problems. It has some noise, and 1/40th is simply too slow for even minor tremors and movement in the subject, but i don't mind the mood. It may be suited to monochrome processing.

View attachment 884886
Did you try the near/far?
 

saladin

Mu-43 Hall of Famer
Joined
May 29, 2015
Messages
2,254
Location
Melbourne
Real Name
jason
Did you try the near/far?


No, i didn't. I can't recall if i have that set , i'll have to check. It may explain the trouble.

Edit: Nope, neither near nor far was set, only AF-on. Something to trial next time.

I'm currently trialling the G9 with a possible shift to the D500 in mind for action shooting. Partly due to optical viewfinder, partly due to AF ability. I can't afford to just jump into the D500 as an addition, so i'd have to sell something off. Which i'm reluctant to do, lol.
 
Last edited:

Brownie

Mu-43 Hall of Famer
Joined
Sep 3, 2018
Messages
3,572
Location
SE Michigan
Real Name
Tim
I recently set my G9 to this function and it works beautifully.
Yes, it's very useful. I have them set to the function buttons next to the lens, near on top and far on the bottom. I can reach them with my middle and ring fingers without having to reset my grip position. The problem is half the time I forget it's there until it's too late!
 

CyVan

Mu-43 All-Pro
Joined
Mar 9, 2016
Messages
1,346
Location
Jamaica
I recently set my G9 to this function and it works beautifully.
Do you mean you make use of them? its a instant use button press not a mode that the camera goes into.
I only ask because a surprising number of people still think its a mode that you put the camera in because of Panasonic's poor documentation.
 

saladin

Mu-43 Hall of Famer
Joined
May 29, 2015
Messages
2,254
Location
Melbourne
Real Name
jason
I like that idea. I might set it up for a specific custom mode position with other appropriate settings for "sport" .
 

grcolts

Mu-43 Top Veteran
Joined
Feb 1, 2010
Messages
511
Location
Texas
Real Name
Gary
In the menu you can set or assign the back/forward function to any button. Yes, it is an instant use of the button and works very well too.
Do you mean you make use of them? its a instant use button press not a mode that the camera goes into.
I only ask because a surprising number of people still think its a mode that you put the camera in because of Panasonic's poor document
 
Links on this page may be to our affiliates. Sales through affiliate links may benefit this site.
Mu-43 is a fan site and not associated with Olympus, Panasonic, or other manufacturers mentioned on this site.
Forum post reactions by Twemoji: https://github.com/twitter/twemoji
Forum GIFs powered by GIPHY: https://giphy.com/
Copyright © Amin Forums, LLC
Top Bottom