Posted by Larry Fischer on January 8, 2009 at 1:59 PM
OK, I see all these posts with "HDR".
I sorry to say I don't know what that is.
On the Nikonian site they showed, under the heading HDR, a 3-frame photo with a Nikon D300 with normal exposure, one 1/2 stop up, and one with 1/2 stop down.
The reply indicated that with 1/2 stop there should have been more photos because the benefit to obtain HDR ofte is better with a wider spread of exposure (if I even understood that right!).
My primal interpretation of HDR was a video term since I've seen it used most converstations about SLR video.
Help and thanks.
Dumb Larry
Posted by Keith Bauer on January 8, 2009 at 2:21 PM
Hi Larry:
HDR stands for High Dynamic Range. It is a technique to capture more dynamic range in a finished image than is possible in a single exposure. That basic technique it to take multiple frames (at least 3) of a scene with exposures varying from some number of stops underexposed to some number of stops overexposed based on the metered value. Then either blend in Photoshop, or use software such as Photomatix Pro to create an image that takes the "properly" exposed parts of each and creates a finished product.
Posted by Dan Borzynski on January 8, 2009 at 3:01 PM
Keith's answer is a good summary.
One problem I frequently see with HDR is a lack of shadows or contrast. When you look at a picture, the darker regions often give the viewer a clue as to what the important pieces of the image is. Once you get rid of these clues, the eye loses its natural cues as to where to look.
HDR is a very effective technique for bringing out detail in shadows and highlights but it needs to be used carefully. Some people do a very good job with it (especially many of the more experienced PS contributors here) but I've also seen shots that just don't look "right".
In a way, it is like the USM tool. The first couple of times you use it, the result is a mess. Then you get better at it and eventually the results become fantastic.
(Disclaimer: This is not a knock on using HDR. Like any of the PS tools, it just takes some experience to use it to its best advantage.)
(Second disclaimer: I've also seen it used "incorrectly" for artist interpretation of a scene to great effect.)
Posted by Larry Fischer on January 8, 2009 at 5:15 PM
I had thought of High Dynamic Range.
So, another question: A sensor will only gather light over a certain range, 0-255, right?
I guess the dynamic range of a sensor is subject to bit depth? 8, 12, 14, etc. If so, which is more important, multiple exposure or bit depth?
Changing exposure won't make 0 less than 0 or 255 more than 255.
So, taking multiple exposure just fills the histogram more fully? Then what about blowouts in the white range when trying to seek shadows???
What is the difference between HDR and D-Lighting in Nikon's NX2? Other programs with similar processes?
My understanding of NX2 is that is uses a different alogrhythm to compute the photo. The histogram does move but does not appear to become "fuller" or more exapansive between 0 and 255.
Will be leaving my computer until Monday but I look forward to some neat presentations on my return. Hopefully, after reading them, I will not be Dumb Larry anymore.....
Thanks again.
Posted by David Bostedo on January 8, 2009 at 5:34 PM
Larry - The dynamic range of a sensor is a physical property of the sensor, and has nothing to do with bit depth, the 0-255 range, or anything else. There is a certain minimum signal that a pixel will always record, even with the lens cap on. There is also a certain maximum amount of light a pixel can record, before the pixel can't reach a higher light level. The difference is the dynamic range.
The 0-255 thing doesn't directly have to do with the dynamic range. Those values are completely arbitrary. Assume that over a certain time (your shutter speed) part of the sensor is being hit by the brightest light it can record without being saturated, and part is being hit by as little light as it can record. We arbitrarily assign these two amounts of light to 0 and 255. It could just as easily be 0 and 400000. Anything darker is automatically 0. And anything brighter is automatically 255.
The reason you have 0 - 255 is because that's how many values can be represented in 8-bits of binary computer code. In a 16-bit image, the values are actually 0-65535. The lightest and darkest amounts of light don't change - only how they are represented to the computer. Each "step" from one level to the next is much smaller with 65535 steps, instead of 255 steps.
This is somewhat related to dynamic range in that you want to pick as many steps as you can and still have them be useful. Each pixel value on your sensor is only somewhat accurate - it might be a little high, it might be a little low. This is due to random noise of the pixel, and related circuitry. Because of this, if you split your values into tiny little steps, you won't see any improvement in the image, because the noise will still be making you pick a random step amongst the ones available. The larger your dynamic range is, and the lower the noise is, the more steps you can have, and still have them be useful steps.
Usually, the way things get encoded, 1 bit is used for each stop of light. (Each bit you add doubles the number of levels, and each stop of range double the amount of light that can be recorded.) So cameras with 14 bit capture should have 2 more stops of dynamic range than those with 12 bits. Currently, only the D3 (although I haven't heard about some of the newer cameras, like D3x or 5DMkII) really gets any real use out of 14 bits - and only a small benefit in very specific circumstances. With others, 14 bits is a bit of a marketing ploy. They've added two bits, but the noise makes those two bit values at the end completely random.
Posted by David Bostedo on January 8, 2009 at 5:43 PM
Changing exposure won't make 0 less than 0 or 255 more than 255.
Correct. But to be clear, a higher exposure (longer shutter, wider aperture, higher ISO) can allow more pixels to collect more light and get closer to the 255 point. Likewise, a lower exposure and cause more pixels to collect less light, moving their values closer to 0.
So, taking multiple exposure just fills the histogram more fully? Then what about blowouts in the white range when trying to seek shadows???
This is pretty oversimplified, but when doing HDR, you only use the "good" portions of each image. So when you take an image for shadow detail, large portions might be blown out. And when you take an image for highlights, large portions may be pure black, but you'll take those parts of the final image from each other. So you won't necessarily have any portions of the image pure white or pure black. (Google HDR and there are tons of demos, tutorials, etc.)
What is the difference between HDR and D-Lighting in Nikon's NX2? Other programs with similar processes?
These types of features just tweak two things : the way a single frames exposure is calculated, and the way the conversion from RAW to image file is handled. They don't actually let you exceed the maximum dynamic range of the sensor. Usually, they just force the camera to underexpose, and then apply a curve to make the shadows lighter.
My understanding of NX2 is that is uses a different alogrhythm to compute the photo. The histogram does move but does not appear to become "fuller" or more exapansive between 0 and 255.
There are lots of different algorithms to convert RAW files to JPG or TIFF. It's not actually that straight forward at all. But HDR doesn't mean you're trying to "fill the histogram". Pretend you are trying to take a picture of a black box on the ground just below the sun. You'll need a VERY short exposure to capture a decent sun, and a long exposure to get texture detail on the box. In the first exposure, most things are dark (except the sun). In the second most things are white (except the box). Combining these would yield an HDR image that would do something your camera couldn't - capture a non-blown out sun, and texture in the box in the same frame. The histogram, however, would probably be spiky at the ends, and not "full".
Posted by Robert Fisher on January 9, 2009 at 11:05 AM
There's a lot of pretty good 'technical' information here but let's simplify it a bit.
Larry, your camera's sensor can capture from full black to full white. It's how much in between it can capture that determines the dynamic range of the sensor. This is basically what David was getting at. Let's look at a simple example. Take an outdoor scene with a brightness range of 9 stops. Let's say your camera, at its best, is only able to capture 6 1/2 stops of brightness. Your histogram may still be 'full' but you're not able to capture the full range of the scene. You have a decision to make. Do you want to sacrifice shadow detail or highlight detail?
HDR is a tool that can help you eliminate that decision. Using HDR (and blended exposures before there was HDR) you can 'expand' the dynamic range of your camera by taking bracketed exposures and combining them after the fact.
Now, there are no monitors (currently) and no printers (currently) that can display an HDR image. If you look at an HDR image on your screen you'll see a lot of dark (black) and a lot of bright (white) because the dynamic range 'created' by the software is too great to be properly shown. In order to generate an image that is viewable on screen and printable the expanded drange of the HDR image has to be compressed. That process of compression is called tonemapping. Tonemapping is a process you go through that takes your HDR image and brings it back into an LDR (Low Dynamic Range) space. It's that process of compressing the brightness values into a smaller space that causes the lack of contrast that someone else mentioned and that causes images to look 'off' to use an earlier respondents term. It can result in very muddy colours and very dull, lifeless photos. I'm not going to go into the use of tonemapping to generate the whacked out, comic book look that is so popular right now.
Tonemapping is a skill. Just as using other tools in digital editing are skills. It takes practice and practice and practice. With practice and judicious use of the tonemapping tools, you can generate an image that has a wider drange than you could capture with your sensor in a single shot but that still has decent contrast so it looks good to the eye - photographically realistic.
Tonemapping isn't, for many people, the last step either; but rather just the start. Once they have their tonemapped image that becomes the basis for other edits - Curves, Levels, HSB, and the like - in PS. With a well made and well edited HDR image, it won't be obvious at all that you're looking at an HDR image.
Posted by Larry Fischer on January 12, 2009 at 9:55 AM
Wow, I now have a lot to chew upon.
I really appreciate both the very technical and simplified versions.
I do a lot of low light photography and need to expand my abilities the the extent possible in this arena.
I'm going to print this page and put it in my photo primer that I keep.
Additionally I will do more study into HDR as it seems an avenue to increase the depth of my photo knowledge base. Google, here I come.
One last question: Doing low light, high contrast photos (sunsets, rises, moon) my Nikon D300 has a bracketing feature. I've not used it but if I do, and overlay the resulting photo, is that a quick way to begin the HDR process?
I think I understand the more photos done with different exposures and combined the better the result at expanding the dynamic range, but at some point to the expense of other areas such as color and crispness??.. The better the technique the better the result but there are still limitations. Comic book color, nope, not for me.
Thank you very much. the fisch
Posted by David Bostedo on January 12, 2009 at 9:59 AM
Doing low light, high contrast photos (sunsets, rises, moon) my Nikon D300 has a bracketing feature. I've not used it but if I do, and overlay the resulting photo, is that a quick way to begin the HDR process?
Yes Larry...that's a great start and you'll learn a lot trying to blend the images together (look online for techniques). It's not "true" HDR, since you don't go through the technical steps that create an HDR file, but it serves the same basic purpose, and can be very useful.
Discussion
8 repliesI sorry to say I don't know what that is.
On the Nikonian site they showed, under the heading HDR, a 3-frame photo with a Nikon D300 with normal exposure, one 1/2 stop up, and one with 1/2 stop down.
The reply indicated that with 1/2 stop there should have been more photos because the benefit to obtain HDR ofte is better with a wider spread of exposure (if I even understood that right!).
My primal interpretation of HDR was a video term since I've seen it used most converstations about SLR video.
Help and thanks.
Dumb Larry
HDR stands for High Dynamic Range. It is a technique to capture more dynamic range in a finished image than is possible in a single exposure. That basic technique it to take multiple frames (at least 3) of a scene with exposures varying from some number of stops underexposed to some number of stops overexposed based on the metered value. Then either blend in Photoshop, or use software such as Photomatix Pro to create an image that takes the "properly" exposed parts of each and creates a finished product.
One problem I frequently see with HDR is a lack of shadows or contrast. When you look at a picture, the darker regions often give the viewer a clue as to what the important pieces of the image is. Once you get rid of these clues, the eye loses its natural cues as to where to look.
HDR is a very effective technique for bringing out detail in shadows and highlights but it needs to be used carefully. Some people do a very good job with it (especially many of the more experienced PS contributors here) but I've also seen shots that just don't look "right".
In a way, it is like the USM tool. The first couple of times you use it, the result is a mess. Then you get better at it and eventually the results become fantastic.
(Disclaimer: This is not a knock on using HDR. Like any of the PS tools, it just takes some experience to use it to its best advantage.)
(Second disclaimer: I've also seen it used "incorrectly" for artist interpretation of a scene to great effect.)
So, another question: A sensor will only gather light over a certain range, 0-255, right?
I guess the dynamic range of a sensor is subject to bit depth? 8, 12, 14, etc. If so, which is more important, multiple exposure or bit depth?
Changing exposure won't make 0 less than 0 or 255 more than 255.
So, taking multiple exposure just fills the histogram more fully? Then what about blowouts in the white range when trying to seek shadows???
What is the difference between HDR and D-Lighting in Nikon's NX2? Other programs with similar processes?
My understanding of NX2 is that is uses a different alogrhythm to compute the photo. The histogram does move but does not appear to become "fuller" or more exapansive between 0 and 255.
Will be leaving my computer until Monday but I look forward to some neat presentations on my return. Hopefully, after reading them, I will not be Dumb Larry anymore.....
Thanks again.
The 0-255 thing doesn't directly have to do with the dynamic range. Those values are completely arbitrary. Assume that over a certain time (your shutter speed) part of the sensor is being hit by the brightest light it can record without being saturated, and part is being hit by as little light as it can record. We arbitrarily assign these two amounts of light to 0 and 255. It could just as easily be 0 and 400000. Anything darker is automatically 0. And anything brighter is automatically 255.
The reason you have 0 - 255 is because that's how many values can be represented in 8-bits of binary computer code. In a 16-bit image, the values are actually 0-65535. The lightest and darkest amounts of light don't change - only how they are represented to the computer. Each "step" from one level to the next is much smaller with 65535 steps, instead of 255 steps.
This is somewhat related to dynamic range in that you want to pick as many steps as you can and still have them be useful. Each pixel value on your sensor is only somewhat accurate - it might be a little high, it might be a little low. This is due to random noise of the pixel, and related circuitry. Because of this, if you split your values into tiny little steps, you won't see any improvement in the image, because the noise will still be making you pick a random step amongst the ones available. The larger your dynamic range is, and the lower the noise is, the more steps you can have, and still have them be useful steps.
Usually, the way things get encoded, 1 bit is used for each stop of light. (Each bit you add doubles the number of levels, and each stop of range double the amount of light that can be recorded.) So cameras with 14 bit capture should have 2 more stops of dynamic range than those with 12 bits. Currently, only the D3 (although I haven't heard about some of the newer cameras, like D3x or 5DMkII) really gets any real use out of 14 bits - and only a small benefit in very specific circumstances. With others, 14 bits is a bit of a marketing ploy. They've added two bits, but the noise makes those two bit values at the end completely random.
Correct. But to be clear, a higher exposure (longer shutter, wider aperture, higher ISO) can allow more pixels to collect more light and get closer to the 255 point. Likewise, a lower exposure and cause more pixels to collect less light, moving their values closer to 0.
This is pretty oversimplified, but when doing HDR, you only use the "good" portions of each image. So when you take an image for shadow detail, large portions might be blown out. And when you take an image for highlights, large portions may be pure black, but you'll take those parts of the final image from each other. So you won't necessarily have any portions of the image pure white or pure black. (Google HDR and there are tons of demos, tutorials, etc.)
These types of features just tweak two things : the way a single frames exposure is calculated, and the way the conversion from RAW to image file is handled. They don't actually let you exceed the maximum dynamic range of the sensor. Usually, they just force the camera to underexpose, and then apply a curve to make the shadows lighter.
There are lots of different algorithms to convert RAW files to JPG or TIFF. It's not actually that straight forward at all. But HDR doesn't mean you're trying to "fill the histogram". Pretend you are trying to take a picture of a black box on the ground just below the sun. You'll need a VERY short exposure to capture a decent sun, and a long exposure to get texture detail on the box. In the first exposure, most things are dark (except the sun). In the second most things are white (except the box). Combining these would yield an HDR image that would do something your camera couldn't - capture a non-blown out sun, and texture in the box in the same frame. The histogram, however, would probably be spiky at the ends, and not "full".
Larry, your camera's sensor can capture from full black to full white. It's how much in between it can capture that determines the dynamic range of the sensor. This is basically what David was getting at. Let's look at a simple example. Take an outdoor scene with a brightness range of 9 stops. Let's say your camera, at its best, is only able to capture 6 1/2 stops of brightness. Your histogram may still be 'full' but you're not able to capture the full range of the scene. You have a decision to make. Do you want to sacrifice shadow detail or highlight detail?
HDR is a tool that can help you eliminate that decision. Using HDR (and blended exposures before there was HDR) you can 'expand' the dynamic range of your camera by taking bracketed exposures and combining them after the fact.
Now, there are no monitors (currently) and no printers (currently) that can display an HDR image. If you look at an HDR image on your screen you'll see a lot of dark (black) and a lot of bright (white) because the dynamic range 'created' by the software is too great to be properly shown. In order to generate an image that is viewable on screen and printable the expanded drange of the HDR image has to be compressed. That process of compression is called tonemapping. Tonemapping is a process you go through that takes your HDR image and brings it back into an LDR (Low Dynamic Range) space. It's that process of compressing the brightness values into a smaller space that causes the lack of contrast that someone else mentioned and that causes images to look 'off' to use an earlier respondents term. It can result in very muddy colours and very dull, lifeless photos. I'm not going to go into the use of tonemapping to generate the whacked out, comic book look that is so popular right now.
Tonemapping is a skill. Just as using other tools in digital editing are skills. It takes practice and practice and practice. With practice and judicious use of the tonemapping tools, you can generate an image that has a wider drange than you could capture with your sensor in a single shot but that still has decent contrast so it looks good to the eye - photographically realistic.
Tonemapping isn't, for many people, the last step either; but rather just the start. Once they have their tonemapped image that becomes the basis for other edits - Curves, Levels, HSB, and the like - in PS. With a well made and well edited HDR image, it won't be obvious at all that you're looking at an HDR image.
I really appreciate both the very technical and simplified versions.
I do a lot of low light photography and need to expand my abilities the the extent possible in this arena.
I'm going to print this page and put it in my photo primer that I keep.
Additionally I will do more study into HDR as it seems an avenue to increase the depth of my photo knowledge base. Google, here I come.
One last question: Doing low light, high contrast photos (sunsets, rises, moon) my Nikon D300 has a bracketing feature. I've not used it but if I do, and overlay the resulting photo, is that a quick way to begin the HDR process?
I think I understand the more photos done with different exposures and combined the better the result at expanding the dynamic range, but at some point to the expense of other areas such as color and crispness??.. The better the technique the better the result but there are still limitations. Comic book color, nope, not for me.
Thank you very much. the fisch
Yes Larry...that's a great start and you'll learn a lot trying to blend the images together (look online for techniques). It's not "true" HDR, since you don't go through the technical steps that create an HDR file, but it serves the same basic purpose, and can be very useful.