Is HDR really DRC?

My impresion of some HDR processed images is that the resulting image has undergone dynamic range compression (DRC). Some of the less well executed examples seem to lack "brightness". The principle that if a little bit is good, less might even be better seems to apply here.

I am enjoying the HDR threads very much. Its great to see photographers caring and sharing their experiences in trying to produce better images. Although I tend to be a lumper, there seems to be a growing need for an HDR forum. This is not retouching. Its a technique that I predict will grow in popularity until it becomes a common in-camera process. Then we can lament about which manufacturers over do it or don't have it ;-)

Jim

As I understand it, HDR imaging is all about capturing a scene with a high dynamic range and translating it to a format with a much smaller DR (usually 8 bits, for a JPEG image).

Obviously, that has to involve some range compression or range mapping, if you prefer that term. The real trick in creating HDR images is in using more than one mapping curve, and in identifying the areas to apply the different curves.

If you do it right, you can get an image that closely mimics what someone would see looking at the scene in real life. If you screw it up, the brighter or darker sections of the image will look like they're pasted onto each other.

I've thought that "High Dynamic Range" is a bit of a misnomer for this technique. "Compressed Dynamic Range" seems a more truthful description of the results, although I applaud attempts to render images nearer to the way the eye perceives them.
--

Mark Bessey said:

'I don't take snaps - I paint with light' - Tony Hancock

Click to expand...

JPEG doesn't have a dynamic range, just as JPEG doesn't have any reason to include DPI numbers. It is a physical concept unrelated to the numbers stored as pixels.

The visible dynamic range that you put INTO a pixel image is related to the input sensor and the light that hits it. It's stronger when there's areas that are intensely lit and also areas that are not intensely lit. (The sensor may saturate or have a low S/N, limiting the usefulness of that sensor to scenes that have a small physical dynamic range.)

The visible dynamic range that you get FROM a pixel image is related to the output media. It's stronger on good projectors in a dark room because they can give a deep dark black and a bright white.

Everything in the JPEG is numbers. The number of bits has no bearing on dynamic range, but on the number of STEPS from darkest (however dark that is) to lightest (however light that is).

--
[ e d @ h a l l e y . c c ] http://www.halley.cc/pix/

Ed Halley said:

JPEG doesn't have a dynamic range,

Click to expand...

Yes, it does. http://en.wikipedia.org/wiki/Dynamic_range

"Dynamic range is a term used frequently in numerous fields to describe the ratio between the smallest and largest possible values of a changeable quantity"

JPEG has a dynamic range of 256 per color.

This dynamic range has nothing to do with the data that is stored in this dynamic range.

You can take a sensor that has a dynamic range of 2, and store its data in a JPEG. The JPEG file format will still have a dynamic range of 256 per color, although it will only be storing data that has a dynamic range of 2.

I can take an image with a dynamic range of 72dB, compress it into a JPEG that has a dynamic range of 48dB, and then display it on my monitor that has a dynamic range of 60dB.

Clint Sanders said:

Yes, it does. http://en.wikipedia.org/wiki/Dynamic_range

"Dynamic range is a term used frequently in numerous fields to
describe the ratio between the smallest and largest possible values
of a changeable quantity"

Click to expand...

I could edit Wikipedia to say that "dynamic range" is used to describe the phase of the moon, but that wouldn't make it right either. In this case, though, you're misreading it. Wikipedia is just fine but you're assuming it's concerned with the quantum step size or sampling. A simple gut check would point out the error: 255 (the largest possible value) divided by 0 (the smallest possible value) is uncomputable. In the real world, DR is meaningful: 9eV / 2eV = 7eV DR.

DR is how much signal is found at the top versus how much signal is at the bottom. It has nothing to do with how small the quantum differentiation is between those two extremes. Decimal integer values in the JPEG are quantum: you're describing the difference between 255:0 and 1:0 and saying that's in the neighborhood of 256:1 DR.

Another analogy: If a robot drew a 64bit graphic onto gray cardboard with a gray crayon, there may be 2^64 measurable gradations of wax buildup, but the dynamic range would be how bright the gray paper was versus how bright the gray wax was, which is hardly visible. The "smallest value" is not zero, it's gray and reflects about half the photons that hit it. The "largest value" is not zero, it's gray and reflects just a tiny fraction more photons above that. Low dynamic range, high quantum fidelity.

--
[ e d @ h a l l e y . c c ] http://www.halley.cc/pix/

Ed Halley said:

I could edit Wikipedia to say that "dynamic range" is used to
describe the phase of the moon, but that wouldn't make it right
either.

Click to expand...

In this case, it is right though.

Ed Halley said:

In this case, though, you're misreading it.

Click to expand...

I fully understand this simple and basic engineering term, and I didn't get its definition from wikipedia. It just link to wiki because it's a handy reference to educate the layman.

Ed Halley said:

just fine but you're assuming it's concerned with the quantum step
size or sampling. A simple gut check would point out the error: 255
(the largest possible value) divided by 0 (the smallest possible
value) is uncomputable.

Click to expand...

Huh? 255/0 approaches infinity. A simple gut check should tell you jpegs don't have an infinite dynamic range.

http://history.sandiego.edu/gen/recording/dynamic.html

"Surface noise disappeared in the digital era with digital "0" and "1" codes independent of any surface material. Sound quality was now determined by "sampling" and "bandwidth." To develop the revolutionary compact disc in 1980, engineers used a computer to measure a sound wave at 44,100 locations on each cycle. Each of these "sampled" locations was defined by a digital "word" 16 bits long, using 65,536 different combinations of "0" and "1" codes to measure the frequency of the sound. Each of these bit codes was stored as a pit in a thin layer of a polycarbonate disc. A laser beam responded to the different sizes of these pits and a computer chip in the CD player reassembled the sound wave from the 16-bit coded samples. Without noise or distortion, the resulting sounds were remarkbly clear within a dynamic range of 96 dB. More powerful computer chips in SDDS movie projectors and DVD audio players of the 1990s created more "bandwidth," or rate of data flow, and therefore clearer and more accurate sounds. The DVD of 1998 sent 24-bit samples at a rate of 9600 kbps, equivalent to a 144 dB dynamic range."

Every time you add a bit, the range of numbers double. Every time you double DR, it increases 6dB.

The 16-bit audio CD has a dynamic range of 96dB (16 * 6).

The 24-bit DVD has a dynamic range of 144dB (24 * 6)

And the 8-bit per color JPEG has a dynamic range of 48dB per color (8 * 6).

Note that they mentioned "without noise or distortion". That's because if there was so much noise that the DR of the signal they captured was only 60dB, the CD would only contain 60dB of dynamic range, even though it is a 96dB container.

Ed Halley said:

Another analogy: If a robot drew a 64bit graphic onto gray cardboard
with a gray crayon, there may be 2^64 measurable gradations of wax
buildup, but the dynamic range would be how bright the gray paper was
versus how bright the gray wax was, which is hardly visible.

Click to expand...

Yeah, that's what I just said. The original image has a DR, the data that stores the image has a DR, and the output device has a DR.

--

'While not exactly disgruntled, he was far from feeling gruntled. He spoke with a certain

what-is-it in his voice, and I could see that, if not actually disgruntled, he was far from
being gruntled.'
~ Pelham Grenville Wodehouse (1881-1975)

Fellow photograhers: It appears that we are in violent agreement that what we are trying to do is re-create the effect of viewing the scene in person by making it easier for our eyes and brain to be tricked. its helpful to have a concept one can use to pull this off.

When I first got my Oly OM-1 in 1972 I was trying to duplicate the colors and values that I saw in person on the slide. I now try to evoke the feeling of the in-person view. Seems to me HDR is a great way to achieve these results in digital media. I hope to see a forum on this.
--
Jim

sorry if I'm being pedantic there.

Yes, the bit depth a camera uses does not indicate its dynamic range. That I agree with.

JPEG does have a dynamic range of 256 per color though. (48db)

I'm just trying to present the whole picture though. Like I said, if the noise of the sound was great enough, the CD will be limited to that noise, even though it can theoretically give 96dB of dynamic range. The same holds true for image sensors.

There's a confusion here between "dynamic range" and what I could call " dynamic resolution". Sometimes this is confusingly noted as "bits per pixel".

JPEG can resolve a range of 256 levels (per color).

A different format could have a range of 1024 levels. And so on.

But black is still black, and white is still white.

A discussion of the lost fuzzy art of the Zone System may be relevant here, but I'm going out for a pizza.