Expose to the right or correct the exposure in camera RAW

[ajmcdo]

I've just been reading the luminous landscape discussion about exposing to the right <http://www.luminous-landscape.com/tutorials/expose-right.shtml> and I wonder what I should do on a daily shooting basis.

How many forum members shoot to the right "in camera" as a routine? How many adjust the exposure settings to the right in the assorted camera RAW software available (I use APS CS)?

After a digital shot the data is digital, so shouldn't the post processing exposure changes in camera RAW be equivalent to shooting to the right in the first place? Or am I missing something, does a right shift to the exposure setting "in camera" do something to the CCD that makes for a better outcome than post processing in camera RAW software?

I find that I quite often make a positive exposure adjustment of .5 or 1.0 in the camera RAW software to improve my shots, would my results be better achieved by shooting to the right as a routine?

I'm interested to hear your opinions.

Tony McDonogh :shock:

[maderito]

I'll try to answer your question ....

For the 10D sensor which captures 12 bits of data per RGB channel, you get 4096 levels of light (per channel). Let's say those those 4096 levels are distributed acrosss 6 exposure zones (approx dynamic range of the CMOS sensor).

Use a modified Ansel Adams system. Call the darkest portions Zone 1 and the lightest Zone 6. Zone 6 has two times the light of Zone 5, and captures (or quantifies) 2048 levels of light. Zone 5 has 1024 levels, Zone 4 - 512, Zone 3 - 256, Zone 2 - 128, and Zone 1 - 64. Thus you have many more light levels to work with in the higher zones, and you can achieve smoother gradations of color and contrast as you remap these levels during image editing.

Suppose you have an image that has only 3 zones. If you set your exposure to capture those zones as Zones 4 to 6, then you have 3584 levels of light intensity per channel to work with. If the same image were captured in Zones 1 to 3, there would only be 448 levels. So it would make sense to capture your image in the higher zones, and then make levels/curves adjustments that re-map your levels to other zones. Further, the claim is made that the signal to noise ratio is better when exposing to the right.

If this concept is correct, you would have to capture the RAW image data with exposure biased to the right. It wouldn't help try to re-create the expose to right bias after that image data is already imprinted on the sensor.

Because 12 bits gives of color information still gives a lot of signal levels, even in the lower zones, I'm not certain that this theory has practical consequences. However, I do find that color corrections and adjustments are harder on underexposed images compared to those exposed to the right, assuming that in neither case do you have blown highlights or blocked shadows.

Note also that once you convert the 12 bit image to 8 bit, you are now working with 256 rather than 4096 levels. If you don't first edit in 16 bit mode, the penalty for not shooting to the right should be even greater. So in principle, a JPEG image from the camera would benefit even more from exposing to the right. (Redo the math on the above 3 zone example using 256 rather than 4096 levels).

As a practical matter, many avoid exposing to the right for fear of blowing highlights in one or more color channels. Some images cover the entire dyanmic range of the sensor and can't be exposed the right. Some settings give you no choice but to underexpose (for instance, when you need a fast shutter speed).

I try to expose to the right, but I don't find many real world senarios where it is feasible to do.

Hope this helps. :)

[dpanicc1]

They also say expose for highlights with digital just like slide film; digital is not too forgiving if the scene is over exposed; film on the other hand provides more latitude. I always try to expose to the right without blowouts as far as possible. And personally I think if a small percentage of the frame is blown out but doesn't detract from the picture, so what. But that's the beauty of RAW, right? As far as why you're adding a stop of exposure--I think is what you said--means, possibly, that you're metering method or focus point is something other than expectation. Hope this helps a little. Dan.

[Joytek]

I do a lot of tripod night shots woth my 10d and have found that especially at long exposures and high isos there is a LOT less noise in the image if I overexpose it and later tone it down in photoshop as a 16 bit tiff. I too read that article and it has changed the way I expose for sure. An over exposed pic that is later tweaked in ps has a lot more snap to it and doesn't show as much posterization in the shadows and dark areas. Since most of my pics are rather high contrast with very bright lights and dark hard shadows due to the street lighting, it is very important to get as much detail as possible for further processing. Also since I change the orange white balance of the sodium lights to an almost pure white (below 2800K) there is a lot of noise brought out this way (mostly blue patchy stuff) and the more over exposed the pic the less this occurs.
my two cents.

hawk.

w. :wink:

[scottbergerphoto]

There is at least one prior thread on the issue of "Exposure to the Right", where various people weighed in.
http://www.photography-on-the.net/forum/showthread.php?t=22723&highlight=expose+right
In my own limited experience, I find it very hard to get sharp pictures with good detail if the raw shot is underexposed. I also have my review on with the histogram. I try to keep the histogram just off the right end, unless of course, there's no white or light tones in the picture.
Scott

[Mark Kemp]

I am going to disagree slightly,

It depends on why you are trying to adjust the exposure. the above comments are correct if it is a very fine tweak to get ultimate quality, but more genrally if you are having trouble with exposure you may want to go under rather than over.

Basically during RAW import or later in the computer you can usually recover something from underexposed sections of the image. There are often shades of dark grey in deep shadows that still contain data and a bit of lightening can pull it back.

But on the other hand white is white is white and if there are any areas over exposed there is no data there at all and you cannot get it back.

So if you are not sure underexpose slightly rather than over expose then any errors should be correctable. Plus you can sometimes get a slightly higher colour saturation in underexposed shots (if that is desirable for the shot in question)

But time permitting, bracketing is better still and then one of the shots should be good.

Finally if you bracket widely on a tripod without changing position you can combine bits of two images make one good one.

[Mikesht]

I'll try to answer your question ....

For the 10D sensor which captures 12 bits of data per RGB channel, you get 4096 levels of light (per channel). Let's say those those 4096 levels are distributed acrosss 6 exposure zones (approx dynamic range of the CMOS sensor).

Use a modified Ansel Adams system. Call the darkest portions Zone 1 and the lightest Zone 6. Zone 6 has two times the light of Zone 5, and captures (or quantifies) 2048 levels of light. Zone 5 has 1024 levels, Zone 4 - 512, Zone 3 - 256, Zone 2 - 128, and Zone 1 - 64. Thus you have many more light levels to work with in the higher zones, and you can achieve smoother gradations of color and contrast as you remap these levels during image editing.

Suppose you have an image that has only 3 zones. If you set your exposure to capture those zones as Zones 4 to 6, then you have 3584 levels of light intensity per channel to work with. If the same image were captured in Zones 1 to 3, there would only be 448 levels. So it would make sense to capture your image in the higher zones, and then make levels/curves adjustments that re-map your levels to other zones. Further, the claim is made that the signal to noise ratio is better when exposing to the right.

If this concept is correct, you would have to capture the RAW image data with exposure biased to the right. It wouldn't help try to re-create the expose to right bias after that image data is already imprinted on the sensor.

Because 12 bits gives of color information still gives a lot of signal levels, even in the lower zones, I'm not certain that this theory has practical consequences. However, I do find that color corrections and adjustments are harder on underexposed images compared to those exposed to the right, assuming that in neither case do you have blown highlights or blocked shadows.

Note also that once you convert the 12 bit image to 8 bit, you are now working with 256 rather than 4096 levels. If you don't first edit in 16 bit mode, the penalty for not shooting to the right should be even greater. So in principle, a JPEG image from the camera would benefit even more from exposing to the right. (Redo the math on the above 3 zone example using 256 rather than 4096 levels).

As a practical matter, many avoid exposing to the right for fear of blowing highlights in one or more color channels. Some images cover the entire dyanmic range of the sensor and can't be exposed the right. Some settings give you no choice but to underexpose (for instance, when you need a fast shutter speed).

I try to expose to the right, but I don't find many real world senarios where it is feasible to do.

Hope this helps. :)

That's a great explanation, thank you! I will steal it and use for references if you dont mind. I dont know if it's practical, but certanly interesting and helps to understand things. On a practical side, I was always told that it's better to underxpose than to overexpose...

[ajmcdo]

Thanks Maderito and others. You've enhanced my understanding!

[EDad]

I understand what's being said here.... I've been pondering it way too much today.

So riddle me this..... why doesn't the file size increase when you shoot to the right. If more information is stored in the right-side zones, then the files should be bigger as they are used, right?

I used a single color surface for my testing (the wall).

Mike

[Scottes]

So riddle me this..... why doesn't the file size increase when you shoot to the right. If more information is stored in the right-side zones, then the files should be bigger as they are used, right?

Whether the value is 0 or 4096, it still gets stored as 12 bits of info in a 16-bit field. Every pixel, whether black or white or something in between, takes 3 16-bit entries in the file (1 entry each for Red, Green, and Blue).

[EDad]

Whether the value is 0 or 4096, it still gets stored as 12 bits of info in a 16-bit field. Every pixel, whether black or white or something in between, takes 3 16-bit entries in the file (1 entry each for Red, Green, and Blue).

True, but since the RAW format is compressed, it does not truly store 3x12 bits/pixel. An unused zone (the far right zone for argument) should undergo very high compression (seeing as it would require very few bytes to completely describe that zone) and that zone represents 1/2 of the data on the CMOS. So a left-shifted file would be smaller than a right-shifted one.

[dtrayers]

I understand what's being said here.... I've been pondering it way too much today.

So riddle me this..... why doesn't the file size increase when you shoot to the right. If more information is stored in the right-side zones, then the files should be bigger as they are used, right?

I used a single color surface for my testing (the wall).

Mike

It isn't that there's more information, it's that the signal to noise ratio is better as you expose to the right.

[dtrayers]

Here's my duplicaiton of Reichmann's experiment. Both these photos were shot on a tripod in manual mode. They're both 100% crops, converted to JPG at 80% quality in CS. Both were at 100 ISO and at f/5.6. I adjusted the shutter speed so the first was exposed so the histogram was just to the right with a little specular highlights blown. The second was exposed 1-1/2 stops under, so the right side of the histogram was just into the center section of the histogram display. I converted using CS-ACR and adjusted the exposure and shadow settings so the histograms spanned the entire range and there was a little blowout on the specular highlights. After conversion the histogram in CS for both were about the same.

Exposed to the right

http://home.comcast.net/~dtrayers/photos/exposeright.jpg


Underexposed 1-1/2 stops

http://home.comcast.net/~dtrayers/photos/underexpose.jpg

You'll notice that the underexposed example is 112Kb while the exposed right is about 75Kb. That's due to the extra noise in the underexposed sample.

I don't always agree with Reichmann and I really don't understand his take on the Sony 828, but with this concept he's spot on.

[CyberDyneSystems]

Even at a high ISO like 800 if you expose "to the right" the noise is really minimal.. it is when the exposure is spot on that the 10D's amazing high ISO clarity truly shines.

[Tom W]

interesting comparison, Dave. I've come across the circumstances when I've needed to underexpose to keep the shutter speed up in available light situations. I guess this indirectly illustrates the value of a faster lens as well.

I'm kind-of happy that I got hold of the 50 as well as the fast 70-200 Sigma.

[Tom W]

Even at a high ISO like 800 if you expose "to the right" the noise is really minimal.. it is when the exposure is spot on that the 10D's amazing high ISO clarity truly shines.

Maybe somebody has already found the answer, but would it be better to expose "to the right" at ISO 800 or expose normal or slightly under at ISO 200 or 400? Perhaps some experimentation is in order.

[CyberDyneSystems]

Well I don't want to drive everyone nuts by dragging out my old fashion show images again,. but they were taken at ISO 800 in jpeg (hadn't learned to cope with raw yet) and they were exposed rather well,..(luck) and they do NOT look at all like ISO 800...

...and ive seen ISO 200 and even 100 shots underexposed that looked a lot noisier.