Pls show examples where increasing brightness in post better than increasing ISO.

Started Oct 1, 2013 | Discussions
Karl Persson Regular Member • Posts: 209
Re: You are welcome

Jack Hogan wrote:

Reilly Diefenbach wrote:

Pushing two stops in LR5 from 1600 to 6400, the 6400 is slightly better. Slightly finer grain, no color blotching (D800e.) So yeah, I'll go with the higher ISO up until the spike climbs the wall, thanks anyway, gentlemen.

Going from ISO 1600 to 6400 is all 'software' on the D800s - i.e. the ISO6400 setting on the D800s is really the ISO1600 setting (pushed two stops digitally in-camera). You can see it clearly here:

But the X-axis is different, as is the slope at lower EVs. And I don´t see how multiplying ISO1600 values by 4 get you to ISO6400 values (in the graph). Please elaborate (I dont use RAWdigger).

On the D800e ISO 6400 raw values are really ISO 1600 raw values times four

gollywop
gollywop Veteran Member • Posts: 8,284
Re: Protecting important highlights

Steen Bay wrote:

Horshack wrote:

The example, pushed 5EV in post. Notice how the image on the television is retained by shooting at ISO 100. If I had shot at ISO 3200 the TV image would have been blown out.

Yes, but like other similar examples, then all it really shows is that it's a bad thing/idea to blow important highlights. That's true if using ETTR at base ISO in good light, and also true if using a higher ISO in low light.

Except that it's not ETTR at higher than base ISO.  But, barring that, yes: clipping is clipping.

-- hide signature --

gollywop

bobn2
bobn2 Forum Pro • Posts: 61,122
Re: Examples

gollywop wrote:

rubank wrote:

gollywop wrote:

rubank wrote:

bobn2 wrote:

rubank wrote:

bobn2 wrote:

rubank wrote:

The simple fact is
that you have to choose: will you sacrifice the highlights or the shadows. With a scene where the DR is beyond what your camera can capture in a single exposure, this choice is of greater importance than what ISO routine you prefer.

Do you want to preserve highlights it is often better to choose underexposure at low ISO, at the cost of shadow colour and noise (esp. banding). If shadows are more important it is often better to choose high ISO, at the cost of highlight rendering.

Heavy shadow lifting commonly leeds to colour shift and banding noise, at least with badly lit scenery where major parts of the image is in deep shadow.
In daylight situations with minor parts of the image in deep shadow, lifting in post is mostly not a problem (unless you have an ISO-full camera).

These are my findings, based on extensive practice and not on science.

The point about knowing the science is that it lets you take out of the loop a lot of the 'extensive practice'.

Extensive practice has a lot of other benefits. You need 10 000 hours of practice (whatever you engage in) to get proficient. Try it.

That's afigure plucked out of the air with no evidence.

Actually not
http://www.wisdomgroup.com/blog/10000-hours-of-practice/

The point is that science provides direction much faster than trial and error.

There is no such contradiction. Most science is based on trial and error.

Well, a lot of engineering is trial and error. And trial and error has indeed helped science. But the major breakthroughs that have led to modern science, special and general relativity and quantum mechanics, are hardly trial and error. These Kuhn-sian revolutions are greatly thought beyond trial.

Even thought-processes can be, and often are, trial and error.
And how would you categorise the search for the Higgs particle...

Difficult. But, seriously, how would you characterize the "knowledge" that the Higgs particle was even worth searching for?

The would have been interesting.

'Let's build a $6Bn machine, just in case something interesting happens when we smash atoms together at huge energies'.

'OK, which sensors should we fit on it?'

'Dunno, see what you can find on ebay, else it might cost $7Bn'

-- hide signature --

Bob

gollywop
gollywop Veteran Member • Posts: 8,284
Re: Examples

bobn2 wrote:

gollywop wrote:

rubank wrote:

gollywop wrote:

rubank wrote:

bobn2 wrote:

rubank wrote:

bobn2 wrote:

rubank wrote:

The simple fact is
that you have to choose: will you sacrifice the highlights or the shadows. With a scene where the DR is beyond what your camera can capture in a single exposure, this choice is of greater importance than what ISO routine you prefer.

Do you want to preserve highlights it is often better to choose underexposure at low ISO, at the cost of shadow colour and noise (esp. banding). If shadows are more important it is often better to choose high ISO, at the cost of highlight rendering.

Heavy shadow lifting commonly leeds to colour shift and banding noise, at least with badly lit scenery where major parts of the image is in deep shadow.
In daylight situations with minor parts of the image in deep shadow, lifting in post is mostly not a problem (unless you have an ISO-full camera).

These are my findings, based on extensive practice and not on science.

The point about knowing the science is that it lets you take out of the loop a lot of the 'extensive practice'.

Extensive practice has a lot of other benefits. You need 10 000 hours of practice (whatever you engage in) to get proficient. Try it.

That's afigure plucked out of the air with no evidence.

Actually not
http://www.wisdomgroup.com/blog/10000-hours-of-practice/

The point is that science provides direction much faster than trial and error.

There is no such contradiction. Most science is based on trial and error.

Well, a lot of engineering is trial and error. And trial and error has indeed helped science. But the major breakthroughs that have led to modern science, special and general relativity and quantum mechanics, are hardly trial and error. These Kuhn-sian revolutions are greatly thought beyond trial.

Even thought-processes can be, and often are, trial and error.
And how would you categorise the search for the Higgs particle...

Difficult. But, seriously, how would you characterize the "knowledge" that the Higgs particle was even worth searching for?

The would have been interesting.

'Let's build a $6Bn machine, just in case something interesting happens when we smash atoms together at huge energies'.

'OK, which sensors should we fit on it?'

'Dunno, see what you can find on ebay, else it might cost $7Bn'

Yeah, but, if you get it on ebay, make sure they bring the machine with them for in-face payment at a local coffee shop.  I found the the need to do it that way by trial and error -- I mean, it's not rocket science, already.

-- hide signature --

gollywop

Horshack Veteran Member • Posts: 7,601
Re: Protecting important highlights

Steen Bay wrote:

Horshack wrote:

The example, pushed 5EV in post. Notice how the image on the television is retained by shooting at ISO 100. If I had shot at ISO 3200 the TV image would have been blown out.

Yes, but like other similar examples, then all it really shows is that it's a bad thing/idea to blow important highlights. That's true if using ETTR at base ISO in good light, and also true if using a higher ISO in low light.

That is the point of this example. The only way to protect the highlights in this image is to shoot at base ISO and then raise the shadows to recreate the midtones. That wouldn't be possible shooting at the nominal ISO 3200 this scene required.

rubank Senior Member • Posts: 1,091
Re: Examples

bobn2 wrote:

gollywop wrote:

rubank wrote:

gollywop wrote:

rubank wrote:

bobn2 wrote:

rubank wrote:

bobn2 wrote:

rubank wrote:

The simple fact is
that you have to choose: will you sacrifice the highlights or the shadows. With a scene where the DR is beyond what your camera can capture in a single exposure, this choice is of greater importance than what ISO routine you prefer.

Do you want to preserve highlights it is often better to choose underexposure at low ISO, at the cost of shadow colour and noise (esp. banding). If shadows are more important it is often better to choose high ISO, at the cost of highlight rendering.

Heavy shadow lifting commonly leeds to colour shift and banding noise, at least with badly lit scenery where major parts of the image is in deep shadow.
In daylight situations with minor parts of the image in deep shadow, lifting in post is mostly not a problem (unless you have an ISO-full camera).

These are my findings, based on extensive practice and not on science.

The point about knowing the science is that it lets you take out of the loop a lot of the 'extensive practice'.

Extensive practice has a lot of other benefits. You need 10 000 hours of practice (whatever you engage in) to get proficient. Try it.

That's afigure plucked out of the air with no evidence.

Actually not
http://www.wisdomgroup.com/blog/10000-hours-of-practice/

The point is that science provides direction much faster than trial and error.

There is no such contradiction. Most science is based on trial and error.

Well, a lot of engineering is trial and error. And trial and error has indeed helped science. But the major breakthroughs that have led to modern science, special and general relativity and quantum mechanics, are hardly trial and error. These Kuhn-sian revolutions are greatly thought beyond trial.

Even thought-processes can be, and often are, trial and error.
And how would you categorise the search for the Higgs particle...

Difficult. But, seriously, how would you characterize the "knowledge" that the Higgs particle was even worth searching for?

The would have been interesting.

'Let's build a $6Bn machine, just in case something interesting happens when we smash atoms together at huge energies'.

If you take out the words "just in case", that´s pretty much it. It´s called "basic research".
And it´s not like the scientists involved payed for it out of their own pockets.

Even if the Higgs particle hadn´t been found, LHC would still be worth the cost and effort.

'OK, which sensors should we fit on it?'

'Dunno, see what you can find on ebay, else it might cost $7Bn'

-- hide signature --

Bob

OP PhilPreston3072 Senior Member • Posts: 2,586
Re: examples where increasing brightness in post is better than increasing ISO.
1

ultimitsu wrote:

PhilPreston3072 wrote:

Put it this way, did the poster do all he possibly can to preserve highlights when using high ISO? If he did not then I don't know why he's complaining about lost highlight detail. When it comes to lowlight shooting, highlights are much easier to preserve than deep shadows.

When high iso is chosen by the metering system, it is never about high light preservation, it is about overall brightness of the resulting jpeg having "proper brightness" - which is 18% grey.

What you are saying is that people shouldn't do that, they should save highlight using lower iso and brighten the image in processing - that is exactly what "shoot low iso then brighten in post" is all about. You understand the need, obviously. But you insist on arguing otherwise with this thread merely because other people have said it first. It seems narcissistic to me.

All along I've been saying the ISO6400 shot wasn't given a fair chance.  You don't go blowing highlights in raw and then cry out there's no highlight detail.  The ISO6400 shot was doomed to fail.  Some people look at that example and think it's a new commandment to never use high ISO.

 PhilPreston3072's gear list:PhilPreston3072's gear list
Canon EOS 450D Canon EF-S 18-200mm f/3.5-5.6 IS +1 more
Iliah Borg Forum Pro • Posts: 25,750
Re: You might be right, you might be wrong

But it is very hard to deduct from the actual output what is hardware induced and what is software induced.

Actually it is not hard, but if you know it is software-induced, are you going to abandon the familiar and convenient workflow?

The result, as it is with the film, is a combination of the properties of the sensitive material and development. What good is it if I can process 400ASA film so that the grain is smaller and acutance is better?

Iliah Borg Forum Pro • Posts: 25,750
Re: You are welcome

Karl Persson wrote:

Jack Hogan wrote:

Reilly Diefenbach wrote:

Pushing two stops in LR5 from 1600 to 6400, the 6400 is slightly better. Slightly finer grain, no color blotching (D800e.) So yeah, I'll go with the higher ISO up until the spike climbs the wall, thanks anyway, gentlemen.

Going from ISO 1600 to 6400 is all 'software' on the D800s - i.e. the ISO6400 setting on the D800s is really the ISO1600 setting (pushed two stops digitally in-camera). You can see it clearly here:

But the X-axis is different, as is the slope at lower EVs. And I don´t see how multiplying ISO1600 values by 4 get you to ISO6400 values (in the graph).

I may be misunderstanding your question, but here is what happens. If you have X=1 at ISO 1600 it is recorded as "4" for ISO 6400; 2 as 8, 3 as 12, and as a result the histogram contains gaps on the "X" axis. The number of pixels (the height, "Y") that was at the level "1" at ISO 1600 is now should be at the level "4", and so on.

Please elaborate (I dont use RAWdigger).

On the D800e ISO 6400 raw values are really ISO 1600 raw values times four

Jack Hogan Veteran Member • Posts: 6,982
Re: You are welcome

Karl Persson wrote:

Jack Hogan wrote:

But the X-axis is different, as is the slope at lower EVs. And I don´t see how multiplying ISO1600 values by 4 get you to ISO6400 values (in the graph). Please elaborate (I dont use RAWdigger).

On the D800e ISO 6400 raw values are really ISO 1600 raw values times four

Hi Karl,

Those are histograms of the bottom couple of hundred levels of the raw data from four D800 DPR studio scene captures at different exposure (hence the different slopes and y-axis but that is not what that image is there for).

Doubling ISO aotbe results in an approximate doubling of the mean raw values. As long as the amplification is done on the analog signal from the sensor before A/D conversion every integer value after amplification will be filled in because the signal fed to the ADC appears continuous as a result of igain and dithering - much like raising the volume on your radio. The histogram would be full, as it is at ISO800 and ISO1600ish*.

If on the other hand the signal from the sensor is held constant and doubling is applied after it has already been converted to a set of digital integers, the resulting raw values would look something like this:

10,11,12,13,14 at ISO1600 would become 20,22,24,26,28 (every other value present) at ISO3200 and 40,44,48,52,56 (every forth value present) at ISO6400.

It is just a straight multiplication of the ISO1600 integers and gaps will result as shown in the raw histograms above. It would also make no difference whether you applied the multiplication in-camera (by raising the ISO) or in-computer (by moving the 'Exposure Compensation' slider to the right in a well behaved raw convertr) because the signal from the sensor is not involved - you are just acting on the same digital data in both cases. The raw data would simply be the ISO1600 information multiplied by two or four in the example.

Well, in fact there would be a littledifference. What would happen to values 5000,5001,5002,5003 captured at ISO1600 when pushed to 6400 with the same exposure? 20000,20004,20008,20012. Oops, they would be clipped, since saturation for the D800 is around value 16000  But if you shot at the lower ISO the information would be captured - and you may decide that you want to squeeze it into your final image in the end.

So if you are in the field, shoot raw and you have maxed out aperture and exposure time given your artistic constraints (DOF, sharpness, bokeh, blur, detail in highlights) - and you feel the urge to push the ISO to 6400 to make the image pleasantly bright on your D800's screen - keep in mind that you have the option to back down to ISO1600 and gain two additional stops of DR without any IQ penalty whatsoever. Those two additional stops of DR may come in handy in some situation. The downside is that the image may look darker than you are used to on the camera screen and you may have to adjust the capture's brightness to your liking during raw conversion.

Cheers,

Jack

*You guessed it, those missing values in the ISO1600 histogram mean that the D800 stops using analog amplification a little before ISO1600

Iliah Borg Forum Pro • Posts: 25,750
Re: You are welcome

*You guessed it, those missing values in the ISO1600 histogram mean that the D800 stops using analog amplification a little before ISO1600

Not that simple I'm afraid. At 1600 a linearization routine kicks in.

Reilly Diefenbach
Reilly Diefenbach Forum Pro • Posts: 13,298
Re: You are welcome

So far I haven't seen any raw convertor that can push a raw four stops without major color uglies that can't be fixed.  Post some if you have them.  Not interested at all in the graphs.

 Reilly Diefenbach's gear list:Reilly Diefenbach's gear list
Nikon D500 Nikon D850
Horshack Veteran Member • Posts: 7,601
Re: You are welcome

Reilly Diefenbach wrote:

So far I haven't seen any raw convertor that can push a raw four stops without major color uglies that can't be fixed. Post some if you have them. Not interested at all in the graphs.

The D7000/D5100 retains color fidelity very well at +5EV pushes even with the much maligned ACR/LR. Here are base ISO and base ISO -5EV images.

D5100 Base ISO

D5100 Base ISO (exposed -5EV and pushed in LR4)

The D7100 not so much:

D7100 Base ISO

D7100 Base ISO (exposed -5EV and pushed in LR4)

Reilly Diefenbach
Reilly Diefenbach Forum Pro • Posts: 13,298
Re: You are welcome

A good test which proves my point.  The D7100 should never be pushed to that extent, as we all found out to our chagrin.  Had it been shot at the correct ISO, there would have been noise, but not the hideous banding and color pollution.  Had you shot the D5100 at the indicated ISO, there wouldn't be the yellow and purple blotches in the shadows, which cannot be repaired in post. Two stops is about it.

 Reilly Diefenbach's gear list:Reilly Diefenbach's gear list
Nikon D500 Nikon D850
gollywop
gollywop Veteran Member • Posts: 8,284
Re: You are welcome

Horshack wrote:

Reilly Diefenbach wrote:

So far I haven't seen any raw convertor that can push a raw four stops without major color uglies that can't be fixed. Post some if you have them. Not interested at all in the graphs.

The D7000/D5100 retains color fidelity very well at +5EV pushes even with the much maligned ACR/LR. Here are base ISO and base ISO -5EV images.

D5100 Base ISO

D5100 Base ISO (exposed -5EV and pushed in LR4)

The D7100 not so much:

D7100 Base ISO

D7100 Base ISO (exposed -5EV and pushed in LR4)

I'm a bit confused as to the meaning of these shots.  The first camera, according to ExifTool, is indeed a D5100, but the second one comes up as a D5200.

Further, they were all shot with auto WB, which makes any resulting color casts problematic.

-- hide signature --

gollywop

gollywop
gollywop Veteran Member • Posts: 8,284
Re: You are welcome

Reilly Diefenbach wrote:

A good test which proves my point. The D7100 should never be pushed to that extent, as we all found out to our chagrin. Had it been shot at the correct ISO, there would have been noise, but not the hideous banding and color pollution. Had you shot the D5100 at the indicated ISO, there wouldn't be the yellow and purple blotches in the shadows, which cannot be repaired in post. Two stops is about it.

It's not clear at all that it is a D7100. The Exif data indicate it's a D5200.  And all the shots were done with autoWB.

-- hide signature --

gollywop

Reilly Diefenbach
Reilly Diefenbach Forum Pro • Posts: 13,298
Re: You are welcome

It doesn't matter. It's the same sensor with AA filter.

 Reilly Diefenbach's gear list:Reilly Diefenbach's gear list
Nikon D500 Nikon D850
gollywop
gollywop Veteran Member • Posts: 8,284
Re: You are welcome

Reilly Diefenbach wrote:

It doesn't matter. It's the same sensor with AA filter.

Hum. From here

D7100 vs D5200: the 14 key differences (1-4)

1. Sensor: Aliasing vs. non-aliased

Both cameras have a 24-megapixel APS-C sensor measuring 23.5 x 15.6mm. Nikon claims that the sensor in the D7100 is a brand new design, but the key point for most users will be the fact that Nikon’s left out the OLPF (Optical Low Pass Filter), also known as the anti-aliasing filter, in front of the sensor. These are used to prevent moiré (interference effects) with fine patterns and textures, but at the cost of slightly blurring the fine detail. Nikon’s decided that the size and resolution of this sensor makes moiré unlikely and that the gain in sharpness from removing the filter is a bigger benefit. As a result, although the D7100 and D5200 have the same resolution, you can expect the fine detail from the D7100 to be slightly sharper.

Of course, that may be wrong.  I don't have either camera so I don't know for sure.

But beyond all that, I don't think such a test ought be done using auto WB -- even using a tripod, which seems to have moved slightly between the D5200 shots.

-- hide signature --

gollywop

Horshack Veteran Member • Posts: 7,601
Re: You are welcome

gollywop wrote:

I'm a bit confused as to the meaning of these shots. The first camera, according to ExifTool, is indeed a D5100, but the second one comes up as a D5200.

Further, they were all shot with auto WB, which makes any resulting color casts problematic.

D5200 was used as a proxy for the D7100 in the original thread I posted the images to - forgot to mention that here, sorry. WB was adjusted in LR4.

Iliah Borg Forum Pro • Posts: 25,750
Post 2 raws, please

Reilly Diefenbach wrote:

So far I haven't seen any raw convertor that can push a raw four stops without major color uglies that can't be fixed. Post some if you have them. Not interested at all in the graphs.

Post 2 raws, please, same exposure settings, exact same light, exact same scene, 4 ISO speed stops apart.

Keyboard shortcuts:
FForum MMy threads