Nikon D90 HDR (High Dynamic Range) Imaging Test ...
Fletcher
Posts: 3,294 ✭
I don't know what happened to our other thread, but here is a quick HDR imaging test with a coin. The theory is that you shoot three shots using increasing exposure level and combine the three images to make one image containing all of the information. My final image isn't great yet, but it shows the potential for having so much information in a single image. It looks amazing when shooting landscapes so I want to see if it is adaptable to macro coin shots. I just need to work on the process a bit ...
Shots 1 - 3
Final Combined Image
Shots 1 - 3
Final Combined Image
0
Comments
<< <i>The theory is that you shoot three shots using increasing exposure level and combine the three images to make one image >>
OR, you can take one image accurately the first time. Just a thought.
Take a look at the histogram feature of your digital camera, a great way to assess exposure. The HDR feature helps when your scene has bright and dark regions that spread out wider than your 'histogram can'. But if you can get your subject to fit within a single images histogram graph, you do not need to use HDR at all.
So, you are really adjusting contrast 'the hard way'.
If you adjusted contrast on your middle image, you could get an identical result.
In many outdoor shots, you have brightly lit clouds and not so well lit shady trees, this is where the clouds can be 8 or 9 stops above the trees. (A stop is a measure of exposure) But your sensor is only good for, say, six stops of range. No single shot will be able to capture both. You will be full-white or full-black somewhere. Taking one shot to properly expose the clouds...another shot to properly expose the trees...feed it through the HDR...now you have simulated a camera sensor that has a 9 stop range. But you will lose contrast. It is a feature with a specific purpose, mostly outdoor shots or highly shaded shots where you want to preserve details in both the brightest and shaded areas.
I have yet to see a coin which exceeds the dynamic range of my camera. Not close.
You can almost always soften the light to allow a coin to fit into the detector's range, but sometimes that softness is not what I want in an image and I just have to deal with a few blown highlights.
<< <i>...
So, you are really adjusting contrast 'the hard way'.
If you adjusted contrast on your middle image, you could get an identical result.
...
I have yet to see a coin which exceeds the dynamic range of my camera. Not close. >>
That's not entirely accurate, IMO. HDR will typically give you (slightly) better results in terms of contrast than simple contrast enhancement will (has to do with the quantization of the original image and how contrast manipulation works, IIRC). Said a bit differently, the histogram of the contrast-enhanced photo will have "gaps" (particularly if the original photo was in 8 to 14 bits), and the HDR image won't.
That said, I agree with your conclusion that HDR is of limited use to coin photograhpy for the reason you mention (coins typically don't have enough dynamic range to require HDR).
Respectfully...Mike
<< <i>This technique is camera independent and done is software (Photoshop primarily) >>
Oh. I understand. I thought he was doing this with the D90.
<< <i>That's not entirely accurate, IMO. HDR will typically give you (slightly) better results in terms of contrast than simple contrast enhancement will (has to do with the quantization of the original image and how contrast manipulation works, IIRC). Said a bit differently, the histogram of the contrast-enhanced photo will have "gaps", and the HDR image won't. >>
For these digital cameras...and shooting raw...you tend to get a good 12 bits of color data (any bits beyond this tend to be noise...not really useful image data, but good for randomizing that quantization effect). I rarely pull my photos very much, so i'm really only chopping one to two bits off of this during my adjustments. The remaining numeric accuracy is still well beyond what is visible on my printer or monitor. I'm using a pretty good Canon camera sensor. As time goes on, sensors on all brands of SLRs get better with signal-to-noise and produce more bits of real brightness data.
Shooting raw is important for so many reasons. Avoiding those awful 8 bit JPEG images is one of them.
One thing that I find tricky with HDR is that one really must get perfect alignment on all images. That can be a little difficult to do if you have to touch the camera at all to make adjustments between shots (you can configure your camera to perform automatic exposure bracketing to avoid that). The issue is minimized by using a solid bulky camera stand. But if your images going into the HDR algorithm are off by even one pixel, I have noticed some bizarre artifacts in the resultant image and an unusual, sometimes distracting loss in sharpness. If you 'pixel peep' at the result anyway. Shrinking images down for web display...you'll never notice!
I guess i've never run into a HDR situation with a gold coin. And I reallt don't have any of that fantastic red copper to shoot, so i've not experienced the red channel saturation that Mr. Goodman talks about. I've come close on gold, but the sensor on the Canon 5D has a pretty nice wide dynamic range that I cannot complain about. It's the dust on the sensor that bugs me, especially when I use the super-microscopic lens set at it's most microscopic setting. All of the dust resolves clearly and sharply! Ack!
<< <i>Where I seem to get into dynamic range issues is typically with red copper and gold where almost all of the info is in the red channel. I presume that having half as many red pixels on the detector compared to green contributes to this issue. I can generally recover most of the highlights in Camera RAW. >>
Bingo. However, I go about things a bit differently -- I make sure the red channel doesn't get clipped, then jack up the low end to compensate for the underexposed photo that results. The net effect is slightly higher noise as a result of amplifying the low end, but generally you can't see it (particularly if you save in 8bit JPG and/or downsample/resize), and the results (at least to my eyes), look better than trying to bring back the clipped red channel detail on the high end. Said a bit differently, I'd rather lose the low end than the high end.
First of all, any digital camera can be used to produce HDR images. You just need to software (e.g., Photoshop) to do the compositing. For those restricted to 8-bit output, you may need more pictures than the 3 recommended for 12-bit cameras. The main benefit of HDR is to give you adaptive, luminance-sensitive exposure adjustment. This is helpful when you have no control over the lighting of the scene you're photographing and the scene contains a very high dynamic range, typically when a scene contains both indoor and outdoor lighting conditions. Shooting stained glass windows in a cathedral is one example. A perhaps pathological example is standing in a cave and shooting a sunlit snowscape out the opening.
The requirements of the input images is that they all be identically composed, are easily registered onto each other, and are shot with the same aperture to prevent depth of field differences. Beware of auto-bracketing on cameras that will change the aperture when shooting for HDR, unless there are special effects you're going after as a side effect of depth of field changes.
The subject coin, which has a very low dynamic range, isn't a good example for this experiment. A better choice would be something very flashy. The thing is, when shooting coins, it is much easier to benefit from getting the lighting right than from using HDR, because unlike a natural scene, you can control the light.
A "computational photography" technique that might be more suited to coins is superresolution, whereby you take a series of pictures of the same item and use the collection of pictures to generate a picture that contains spatial information that is more precise than any of the input pictures, allowing you to represent the output with higher resolution than any of the inputs. I'll have to try this sometime.
Keeper of the VAM Catalog • Professional Coin Imaging • Prime Number Set • World Coins in Early America • British Trade Dollars • Variety Attribution
<< <i>A "computational photography" technique that might be more suited to coins is superresolution, whereby you take a series of pictures of the same item and use the collection of pictures to generate a picture that contains spatial information that is more precise than any of the input pictures, allowing you to represent the output with higher resolution than any of the inputs. I'll have to try this sometime. >>
IMHO, a set of extension tubes will do a much better job at increasing resolution/magnfication than anything done in post processing.
<< <i>IMHO, a set of extension tubes will do a much better job at increasing resolution/magnfication than anything done in post processing. >>
Unless you've run out of space, can't light the scene, are using a point and shoot, etc. Superresolution techniques may or may not enable you to stretch the limitations of your setup for high magnification applications. It requires testing.
Keeper of the VAM Catalog • Professional Coin Imaging • Prime Number Set • World Coins in Early America • British Trade Dollars • Variety Attribution
<< <i>
<< <i>That's not entirely accurate, IMO. HDR will typically give you (slightly) better results in terms of contrast than simple contrast enhancement will (has to do with the quantization of the original image and how contrast manipulation works, IIRC). Said a bit differently, the histogram of the contrast-enhanced photo will have "gaps", and the HDR image won't. >>
For these digital cameras...and shooting raw...you tend to get a good 12 bits of color data (any bits beyond this tend to be noise...not really useful image data, but good for randomizing that quantization effect). I rarely pull my photos very much, so i'm really only chopping one to two bits off of this during my adjustments. The remaining numeric accuracy is still well beyond what is visible on my printer or monitor. I'm using a pretty good Canon camera sensor. As time goes on, sensors on all brands of SLRs get better with signal-to-noise and produce more bits of real brightness data.
Shooting raw is important for so many reasons. Avoiding those awful 8 bit JPEG images is one of them.
One thing that I find tricky with HDR is that one really must get perfect alignment on all images. That can be a little difficult to do if you have to touch the camera at all to make adjustments between shots (you can configure your camera to perform automatic exposure bracketing to avoid that). The issue is minimized by using a solid bulky camera stand. But if your images going into the HDR algorithm are off by even one pixel, I have noticed some bizarre artifacts in the resultant image and an unusual, sometimes distracting loss in sharpness. If you 'pixel peep' at the result anyway. Shrinking images down for web display...you'll never notice!
I guess i've never run into a HDR situation with a gold coin. And I reallt don't have any of that fantastic red copper to shoot, so i've not experienced the red channel saturation that Mr. Goodman talks about. I've come close on gold, but the sensor on the Canon 5D has a pretty nice wide dynamic range that I cannot complain about. It's the dust on the sensor that bugs me, especially when I use the super-microscopic lens set at it's most microscopic setting. All of the dust resolves clearly and sharply! Ack! >>
Thanks for the response....
Two points. First, you do realize that when you save to JPG you lose all that data above 8 bits, don't you.
Second, there's no reason dust on the sensor should be sharp unless you're using a very high aperature, which would lead to diffraction in the original image to begin with. Also, you might want to investigate sensor cleaning if it is particularly bad and affecting your images.
<< <i>
<< <i>IMHO, a set of extension tubes will do a much better job at increasing resolution/magnfication than anything done in post processing. >>
Unless you've run out of space, can't light the scene, are using a point and shoot, etc. Superresolution techniques may or may not enable you to stretch the limitations of your setup for high magnification applications. It requires testing. >>
Good points. I suppose I'm just spoiled with my DSLR and tend to forget digicams and lighitng challenges.
<< <i>I didn't say it was going to be great ... I'm simply playing around with the technique to see if it could potentially add anything of value to the numismatic imaging process. Being retired affords me the time to tinker
Well, go tinker with a lustrous gold coin.
<< <i>
<< <i>I didn't say it was going to be great ... I'm simply playing around with the technique to see if it could potentially add anything of value to the numismatic imaging process. Being retired affords me the time to tinker
Well, go tinker with a lustrous gold coin.
Werd. Just be careful not to change the lighting or the lens aperture.
Keeper of the VAM Catalog • Professional Coin Imaging • Prime Number Set • World Coins in Early America • British Trade Dollars • Variety Attribution
<< <i>
<< <i>
<< <i>I didn't say it was going to be great ... I'm simply playing around with the technique to see if it could potentially add anything of value to the numismatic imaging process. Being retired affords me the time to tinker
Well, go tinker with a lustrous gold coin.
Werd. Just be careful not to change the lighting or the lens aperture. >>
Or the focus or the camera position.
<< <i>I didn't say it was going to be great ... I'm simply playing around with the technique to see if it could potentially add anything of value to the numismatic imaging process. Being retired affords me the time to tinker
p.s. thanks for posting your experiements. THis is a topic I've wanted to discuss for a while now.
BTW - Using Nikon Camera Control 2 allows me to run my camera directly from my computer and see what I am going to get (using the live view mode) before I take the picture. Amazingly helpful software ...
<< <i>I just picked up a nikon D90, missed FedEx friday so it should be here monday. I can't wait to start playing around with it. What type of lens were you using? A macro lens? >>
Kind of off topic, but when I started with digital photograph I had a Nikon D40 and on a lark I bought a used Kiron/Lester S1 105mm as a supplemental lens. While I had to use it in manual mode, I have to say that when my camera was mounted it produced some of the best pictures I've been able to make - ever.
I'm now using a Canon 450D aka Rebel Xsi, because I'm cheap. I'm happy with it because the Canon software allows me to shoot in live-view and works on OS X (Leopard/10.5.x), which is a huge boon for conveyor belting coin images. But, I'm not terribly pleased with either the 18-55mm kit lens nor the 50mm macro lens. Eventually I'll invest in a better copy stand to give me the working distance I need for the Tamron SP AF 180mm f/3.5 Di LD IF Macro or the Canon EF 100mm f/2.8 USM Macro.
But now I've really drifted
D90 Video Example
Since HDR is on page 1 again, this old thread is informtative.
---------------------------------------------------------------------------------------------------------------------------------------
BONGO HURTLES ALONG THE RAIN SODDEN HIGHWAY OF LIFE ON UNDERINFLATED BALD RETREAD TIRES
and they're cold.
I don't want nobody to shoot me in the foxhole."
Mary
Best Franklin Website
Leo
The more qualities observed in a coin, the more desirable that coin becomes!
My Jefferson Nickel Collection