RAW dynamic range

[OldClicker]

We've had several discussions on the dynamic range advantage (or not) of shooting in RAW. Here is a crude demo on what I mean when I say that RAW will give you a wider dynamic range. I over exposed by +3 an extra-fine jpg and a RAW and then adjusted the exposure by -3 of both using Lightroom 3. You can clearly see that the jpg adjustment does nothing but make what is already there darker. The blown out whites are still blown out. With the RAW, the adjustment recovers all of the information that was in the original scene.


Over exposed in camera jpg:




Over exposed RAW jpg:




jpg with adjusted exposure:




Raw with adjusted exposure:




Terry

[Asmarlak]

It seems like some colors became similar, ran together, and/or lost in the Adjusted RAW and became less defined than in the adjusted JPG.
I think if you use a camera with a good dynamic range to begin with, you won't have the hot white that you see in these JPG's, and you'd capture the best of both worlds (colors and highlights).
I think "RAW or JPEG" will always be discussed because it will always go down to the camera used, just my opinion.
Here is an example of the most difficult lighting any camera has to deal with but my E-30 got it perfectly right in every way.

[armando_m]

Excellent demonstration!

on the recovered raw I start to see a difference in the 3% gray

whereas on the recovered jpg I do not see a difference until 14% gray, maybe at 12 %

[Medley]

My only argument here is with semantics Terry: Dynamic range is the ratio between the largest and smallest possible values of a changeable quantity, such as in sound and light. It is measured as a ratio, or as a base-10 (decibel) or base-2 (doublings, bits or stops) logarithmic value.

Dynamic range is expressed as a ratio between the darkest value that is not black, and the lightest possible value. In the JPEGs then, the dynamic range is 1:255. So what is the dynamic range of the unadjusted RAW image?

Can you find a pixel value in the RAW that is higher than 255? The reason that you can't, is that the monitor can't reproduce the amount of contrast in the HDR image. So by the time you see the image, it already fits within the dynamic range of the monitor.

The dynamic range of the unadjusted RAW image then, is 1:255. The dynamic range of the two images are identical.Black's value is 0, white's is 255.

But your camera is certainly capable of capturing an HDR image, so how did it become LDR?

There are LOTS of ways it CAN be done, but it basically boils down to three main methods:

1) You can reduce the contrast. By lowering the difference in values between the lightest and darkest you can eventually get the image to fit within a lower dynamic range,

2) You can clip the highlights and shadows, and compress the midtones to fit the lower dynamic range. This gives you better overall contrast, but you lose information in the clipping, or

3) You can tone-map the image. Tone-mapping involves looking at the dynamic range of both the HDR image and LDR output target, and determining the new value based upon where the pixel falls on the gamma curve. This keeps the localized contrast between pixels as accurate as possible.

So, realizing that the dynamic range of the two images are identical, and realizing that by the time you were able to view the 'unadjusted' RAW image it had, in fact, already been converted to LDR so that it would display properly on your monitor, to what do you attribute the difference in the images?

- Joe U.

[Franglais]

First - I like Asmarlak's photo. Luminous shadows with lots of detail - I like that sort of stuff. The shot was probably taken early in the morning or late in the afternoon when the lighting contrast was fairly low.

I'm - puzzled about Medley's post. I don't think you're talking about dynamic range at all.

- Each of the three primary colours on a JPG is coded in 8 bits which gives you 255 levels from darkest to lightest
- Each of the three primary colours in a RAW file is coded on 12 bits (or even 14 bits) which gives you 4096 (or 16384) levels

That doesn't mean that a RAW image has a dynamic range of 1:4096. It just means that the signal coming off the sensor is recorded more finely.

When the RAW image is converted into a JPG the conversion process squeezes the 4096 levels into 255 and gives it a pleasing appearance. In a high contrast scene, it tries to render the mid tones with "normal" contrast and the information in the darkest shadows and brightest highlights is clipped - you get a JPG with no information in those zones. With a RAW editor you can reduce the clipping and get that detail back. This usually means giving the tonal curve an S shape like film. Get it wrong and the whole thing looks unnatural.

The dynamic range of a modern DSLR sensor is about 11000:1 or 13 stops according to this article http://en.wikipedia.org/wiki/Dynamic_range#Photography

[OldClicker]

My only argument here is with semantics Terry: Dynamic range is the ratio between the largest and smallest possible values of a changeable quantity, such as in sound and light. It is measured as a ratio, or as a base-10 (decibel) or base-2 (doublings, bits or stops) logarithmic value.

Dynamic range is expressed as a ratio between the darkest value that is not black, and the lightest possible value. In the JPEGs then, the dynamic range is 1:255. So what is the dynamic range of the unadjusted RAW image?

Can you find a pixel value in the RAW that is higher than 255? The reason that you can't, is that the monitor can't reproduce the amount of contrast in the HDR image. So by the time you see the image, it already fits within the dynamic range of the monitor.

The dynamic range of the unadjusted RAW image then, is 1:255. The dynamic range of the two images are identical.Black's value is 0, white's is 255.

But your camera is certainly capable of capturing an HDR image, so how did it become LDR?

There are LOTS of ways it CAN be done, but it basically boils down to three main methods:

1) You can reduce the contrast. By lowering the difference in values between the lightest and darkest you can eventually get the image to fit within a lower dynamic range,

2) You can clip the highlights and shadows, and compress the midtones to fit the lower dynamic range. This gives you better overall contrast, but you lose information in the clipping, or

3) You can tone-map the image. Tone-mapping involves looking at the dynamic range of both the HDR image and LDR output target, and determining the new value based upon where the pixel falls on the gamma curve. This keeps the localized contrast between pixels as accurate as possible.

So, realizing that the dynamic range of the two images are identical, and realizing that by the time you were able to view the 'unadjusted' RAW image it had, in fact, already been converted to LDR so that it would display properly on your monitor, to what do you attribute the difference in the images?

- Joe U.

Interresting, but not contradictory to the post. The point is that the RAW contains a higher dynamic range than the in camera jpg. What use you make of this is up to you, but the information is there now and 10 years from now (when we all have HDR monitors). - Terry

[OldClicker]

It seems like some colors became similar, ran together, and/or lost in the Adjusted RAW and became less defined than in the adjusted JPG.
I think if you use a camera with a good dynamic range to begin with, you won't have the hot white that you see in these JPG's, and you'd capture the best of both worlds (colors and highlights).
I think "RAW or JPEG" will always be discussed because it will always go down to the camera used, just my opinion.
Here is an example of the most difficult lighting any camera has to deal with but my E-30 got it perfectly right in every way.

If the dynamic range fits in the 8-bit file (almost, there are blacks), then this image is not 'the most difficult". Also, if you had the desire to pull the detail out of those shadows in this image, you could not. If it was RAW you could. Any modern camera can capture the dynamic range of an 8-bit jpg. - Terry

[OldClicker]

First - I like Asmarlak's photo. Luminous shadows with lots of detail - I like that sort of stuff. The shot was probably taken early in the morning or late in the afternoon when the lighting contrast was fairly low.

I'm - puzzled about Medley's post. I don't think you're talking about dynamic range at all.

- Each of the three primary colours on a JPG is coded in 8 bits which gives you 255 levels from darkest to lightest
- Each of the three primary colours in a RAW file is coded on 12 bits (or even 14 bits) which gives you 4096 (or 16384) levels

That doesn't mean that a RAW image has a dynamic range of 1:4096. It just means that the signal coming off the sensor is recorded more finely.

When the RAW image is converted into a JPG the conversion process squeezes the 4096 levels into 255 and gives it a pleasing appearance. In a high contrast scene, it tries to render the mid tones with "normal" contrast and the information in the darkest shadows and brightest highlights is clipped - you get a JPG with no information in those zones. With a RAW editor you can reduce the clipping and get that detail back. This usually means giving the tonal curve an S shape like film. Get it wrong and the whole thing looks unnatural.

The dynamic range of a modern DSLR sensor is about 11000:1 or 13 stops according to this article http://en.wikipedia.org/wiki/Dynamic_range#Photography

The only comment that I would have here is from,

"Get it wrong and the whole thing looks unnatural."

It seems to me that 'unnatural' to most photographers actually means unphotographic or unfilm-like - not that it is uncharacteristic of human vision.

Terry

[Medley]

Then explain to me the HDR mode on the Pentax K-7 DSLR which outputs (only) a tone-mapped JPEG file.

And if you're going to tell me that it is in the processing, not the file format, I'm going to tell you that is precisely my point.

Wikipedias definition of dynamic range: http://en.wikipedia.org/wiki/Dynamic_range

Now if you'd like, you can follow links to contrast ratios, and EVs and logarithms and formulas. Then you can find the contrast ratio of your monitor......

You get the drift. Just don't tell me that one of those images has a greater dynamic range than the other. It doesn't. It did at one time, but it was tone-mapped into a lower dynamic range to post here. Yet there IS a difference. So I submit that the difference MUST be in the processing, not in the dynamic range. And the only difference WHERE that processing takes place makes, be it in the camera, or in Photoshop, or in conversion for output, is how much control you, the end user, has over the process.

But isn't that what we're all after- The maximum amount of creative control over our HDR images?

As I said, semantics.

- Joe U.

[OldClicker]

Then explain to me the HDR mode on the Pentax K-7 DSLR which outputs (only) a tone-mapped JPEG file.

And if you're going to tell me that it is in the processing, not the file format, I'm going to tell you that is precisely my point.

Wikipedias definition of dynamic range: http://en.wikipedia.org/wiki/Dynamic_range

Now if you'd like, you can follow links to contrast ratios, and EVs and logarithms and formulas. Then you can find the contrast ratio of your monitor......

You get the drift. Just don't tell me that one of those images has a greater dynamic range than the other. It doesn't. It did at one time, but it was tone-mapped into a lower dynamic range to post here. Yet there IS a difference. So I submit that the difference MUST be in the processing, not in the dynamic range. And the only difference WHERE that processing takes place makes, be it in the camera, or in Photoshop, or in conversion for output, is how much control you, the end user, has over the process.

But isn't that what we're all after- The maximum amount of creative control over our HDR images?

As I said, semantics.

- Joe U.

Who are you replying to? - Terry

[Franglais]

You get the drift. Just don't tell me that one of those images has a greater dynamic range than the other. It doesn't. It did at one time, but it was tone-mapped into a lower dynamic range to post here. Yet there IS a difference. So I submit that the difference MUST be in the processing, not in the dynamic range. And the only difference WHERE that processing takes place makes, be it in the camera, or in Photoshop, or in conversion for output, is how much control you, the end user, has over the process.

.

Still a bit puzzled when you say there IS a difference.

I checked up on Pentax K7 HDR:

When the Pentax K7 does an HDR image it takes 3 pictures in quick succession - normal exposure, -3 stops and +3 stops, then combines them. Most of the final image comes from the normally exposed picture but where highlights are blown in the normal exposure it will blend in detail from the -3 stops picture and for deep shadows it will take detail from the +3 stops picture. The result is a single JPG file with detail all over.

Oldclicker's post demonstrates that if you take a single RAW picture and the highlights are burnt out when you do a conversion to JPG then all is not lost. You can get back the detail with a RAW editor. In most editors you do this by reducing the exposure 1 or 2 stops. This brings back your highlights but it makes the whole picture darker. You have to bring up the midtones with the Curves tool. Lightroom is smart because it has a built-in Recovery button for highlights that does the same thing in one step