Visit MPB Visit MPB Visit MPB

K7 HDR hypothetical question

BODYHEAT
Posted 04/07/2009 - 16:51 Link
As a newcomer to digital SLRs I am going through a learning curve because is some ways it is similar to film and in others not at all. In a very simplistic way one of the major differences is that film is about chemistry (physical things which need manipulation) whereas digital is about software (computer code). (hang in there, I'm leading up to something)

One question has been bugging me ever since I learned about the new HDR feature in the K7 in which it takes 3 bracketed shots (over, under, and well exposed) and digitally combines them into a single shot to increase the dynamic range. And the question is why?

This HDR feature sounds like an exciting development except for the limitation of moving objects (which would include leaves on trees blowing in the wind) or handheld shots due to camera shake. We are told that for this function to work it must be done on a tripod so the shots match up.

If digital raw files can basically capture all the data (read: code) from the image coming through the lens for you to select which parts you wish to enhance/suppress later in PP, why does the the K7 need to take three 'physical' shots at all to combine them? Why can't it take a single raw image and 'create' the HDR combination through software manipulation in camera, thus avoiding any potential camera shake and mismatching of images?

I needed to get it off my chest as it has been bugging me that surely this function could have been achieved by a clever software algorithm manipulating the raw data without the need to take three physical shots and risk camera shake.

Is this a dumb question or is there logic in my thinking?
FILM - Pentax: LX, K2 - 24/2.8, 40/2.8, 50/1.4, 120/2.8, 80-200/4.5, 28-105mm 2.8 macro, AF 280T

DIGITAL - Nikon : D300 - 18-200mm f/3.5-5.6G IF-ED AF-S VR DX
Pentaxophile
Posted 04/07/2009 - 17:10 Link
I guess the difference is that the HDR feature does not require any 'manipulation' to boost the image after capture. Therefore there is no increased noise or loss of detail.

The three shots can also give you more latitude than RAW would have given you - you can pull back detail from images up to about one and a half stops overexposed, but that's the limit, and after that the sensor does not pick up any detail, only a solid white block, which no manipulation can put detail back into.

HDR can potentially give you unlimited exposure latitude. Say if you were taking pictures in a cave. If you expose for the cave walls, the mouth of the cave is going to be overexposed by a lot more than a couple of stops - you will have a white circle with no actual detail recorded in it. HDR can give you perfect exposure for the cave walls and the landscape framed by the entrance of the cave. The only drawback is that this would look completely unnatural!
[link=https://500px.com/will_brealey/[/link]
johnriley
Posted 04/07/2009 - 17:13 Link
Not a dumb question at all.

The type of "HDR" where you use a single exposure and process three or more files from it is possible, from RAW or even from a JPEG, but there is a limit to how wide the dynamic range of the sensor is. You are not capturing all the information that comes through the lens, just about 6 or 7 stops of it.

True HDR needs more than three exposures, but certainly at least three. You would expose at the "correct" reading and also at one stop under and one stop over.

Personally I use 7 exposures, -3, -2, -1, 0, +1, +2, +3 which captures a huge dynamic range of about 13 stops. This is not possible without HDR. When these images are combined in Photomatix you get an enormous dynamic range. This suits church interiors, where the deepest shadows and the brightest windows can all be reproduced.

Of course overdone it looks awful, but with the right subject it can look superb. As you have already observed though, it's not for trees on a windy day!
Best regards, John
Edited by johnriley: 04/07/2009 - 17:14
Mike-P
Posted 04/07/2009 - 17:23 Link
The only time I use HDR these days is when I have a situation where the sky will overexpose or ground underexpose when I am trying to get a landscape shot (bright sunny day in the forest). By taking 3 bracket shots of -1, 0, +2 I can normally combine them to get a decently exposed picture.
yelvertoft
Posted 04/07/2009 - 18:40 Link
There are times when you can't get the whole of the picture "correctly" exposed with a single frame, be it film or digital, there simply isn't enough range available. HDR can have the same effect as using an ND grad on a landscape image, if it's done correctly.

As John has said, church interiors are notoriously difficult. Can't stand a lot of over-processed HDR stuff myself (it's just a matter of taste), but I'd never have been able to get this with a single exposure, no matter how much playing around I'd have done. It's a blend of five exposures.

Comment Image


Duncan
Anvh
Posted 04/07/2009 - 19:41 Link
Here is another question.

What is the dynamic range of the type of media we used today:
so print, monitor... and the rest
Stefan
Comment Image

K10D, K5
DA* 16-50, DA* 50-135, D-FA 100 Macro, DA 40 Ltd, DA 18-55
AF-540FGZ
Edited by Anvh: 04/07/2009 - 19:42
johnriley
Posted 04/07/2009 - 23:33 Link
Very much less than is recorded on the sensor or film, but the key to that is just how we compress the dynamic range to fit within the limitations of our materials.

A colour slide has much more dynamic range than a print, and a monitor will have a similar advantage.

This does not negate what we do with HDR or any other technique as our challenge, often met, is to get all the information onto the final viewing media, just like we did in the darkroom when we dodged and burned our prints.
Best regards, John
RR
Posted 04/07/2009 - 23:50 Link
I'm not sure if it's do-able or if it's currently down to computational power in bodies or whatever but instead of taking 3 seperate shots with the shutter closing each time couldn't 3 seperate exposures be computed from one shutter actuation. Follow me if you can & I'll keep the figures simple so you have the best chance of following me.


ie: you want a hdr image of 3 exposures you take one at -1 ev one normal & one +1 ev

lets say your -1 is a one second exposure, mid is 2 secs, & +1 is 4 secs. Is it not possible that instead of taking 3 exposures of 1,2 & 4 secs that you could take one 4 sec image & the camera could record 3 seperate exposures by "saving the data" at the 1 sec & 2 sec point without using the mechanical shutter. After all it is just recording a load of ones & zeros & should be able to work this out. Digital compacts have no mechanical shutter after all.

Now 4 secs is not going to help with movement but when you scale this to high shutter speeds you could freeze motion in HDR also. Surely it's just a matter of the right algorythm & enough computing power & speed to carry this off ?

Just an idea, do you think I need to patent it & get on Dragons Den ?
Edited by RR: 04/07/2009 - 23:53
Anvh
Posted 05/07/2009 - 00:27 Link
Thank you for the information John still we compress the dynamic range because we can't show any other way I believe.
Wouldn't it be great that they can print a photo with the complete dynamic range that was there, that would be truly stunning.
I know it can't be done but just think about and maybe someday there might be a media that can.

RR I also once thought about that but don't think it is doable with current techniques.
You need to read all the info from all the pixels instantly and that can be done with current design.

Btw compacts do need a shutter but since they don't have one they use the aperture and close it fully so that it acts like a shutter.

The shutter still works like in the film days and it gives the camera time to process the information correctly.
Stefan
Comment Image

K10D, K5
DA* 16-50, DA* 50-135, D-FA 100 Macro, DA 40 Ltd, DA 18-55
AF-540FGZ
Edited by Anvh: 05/07/2009 - 00:30
Don
Posted 05/07/2009 - 02:12 Link
RR wrote:
I'm not sure if it's do-able or if it's currently down to computational power in bodies or whatever but instead of taking 3 seperate shots with the shutter closing each time couldn't 3 seperate exposures be computed from one shutter actuation. Follow me if you can & I'll keep the figures simple so you have the best chance of following me.


ie: you want a hdr image of 3 exposures you take one at -1 ev one normal & one +1 ev

lets say your -1 is a one second exposure, mid is 2 secs, & +1 is 4 secs. Is it not possible that instead of taking 3 exposures of 1,2 & 4 secs that you could take one 4 sec image & the camera could record 3 seperate exposures by "saving the data" at the 1 sec & 2 sec point without using the mechanical shutter. After all it is just recording a load of ones & zeros & should be able to work this out. Digital compacts have no mechanical shutter after all.

Now 4 secs is not going to help with movement but when you scale this to high shutter speeds you could freeze motion in HDR also. Surely it's just a matter of the right algorythm & enough computing power & speed to carry this off ?

Just an idea, do you think I need to patent it & get on Dragons Den ?

I had a similar thought some time back... my idea involved using an lcd panel to mask off the highlights in an image for a portion of the exposure (biased for the dark areas) to capture detial in both highlight and shadow....
but alas it seems most people are simply happy to use flash to control contrast in many situations, and hdr in other situations...

but what your talking about is somewhat like taking a video segment then merging the frames into and hdr still... I suspect that is coming....
Fired many shots. Didn't kill anything.
Edited by Don: 05/07/2009 - 02:14
Anvh
Posted 05/07/2009 - 12:55 Link
RR have been thinking more about it.
Still your idea doesn't seem to work since you need to read the whole sensor out simultaneously, very difficult to do that.

You could however with your example with 1, 2 & 4 second exposure, you could use the mechanical shutter to freeze the sensor for a moment.

You can see a pixel as a bucket that fills itself with light.
So you let the sensor fill for 1 second and close the shutter read out how full they but let them be that full so don't empty them and open the shutter again till the next mark, do you get the idea?
I don't know if the pixel need to be emptied when read though but if not this is quite feasible.
Stefan
Comment Image

K10D, K5
DA* 16-50, DA* 50-135, D-FA 100 Macro, DA 40 Ltd, DA 18-55
AF-540FGZ
womble
Posted 06/07/2009 - 23:05 Link
Anvh wrote:
Here is another question.

What is the dynamic range of the type of media we used today:
so print, monitor... and the rest

From McCollough, Complete guide to digital high dynamic range photography:
Outdoor sunlight: 17 EV
Human eye: 20 EV
Negative film: 10-11 EV
Computer monitor: 9 EV
D-SLR: 6 to 8 EV
High Quality Glossy print: 7.6 EV
Compact digital camera: 6.6 EV
Average quality glossy print: 6.6 EV
Slide film: 6 EV
High quality matte print: 5.6 EV

Some of these may well be out of date now.
Kris Lockyear
It is an illusion that photos are made with the camera… they are made with the eye, heart and head. Henri Cartier-Bresson
Lots of film bodies, a couple of digital ones, too many lenses (mainly older glass) and a Horseman LE 5x4.

My website
Anvh
Posted 06/07/2009 - 23:45 Link
Oh wow thank you Kris that's certainly helpful even if the numbers are old they are a good guide.

Prints will most likely still have the same DR so most D-SLR today will cover that with one shot.
Just was wondering if it was worth taking more photos and merge them together but don't flatten them like we do now and get more details.
I had the idea a monitor had less DR then a print so that was the reason behind it.

I use a CRT monitor so the number would fit that one most likely, TFT will most likely be lower.

Still have my eyes out for a laser tv since they produce 90% of the colours we can see so it would be interested to know what DR that thing has.
The only problem is that there is only one laser tv on the market and it's 65inch and cost $6500, far too big for the computer
Stefan
Comment Image

K10D, K5
DA* 16-50, DA* 50-135, D-FA 100 Macro, DA 40 Ltd, DA 18-55
AF-540FGZ
johnriley
Posted 06/07/2009 - 23:49 - Helpful Comment Link
This of course confirms that slide film has a narrow dynamic range, so if you can expose it properly you won't have a problem with digital.

Slides are a great learning tool and thoroughly recommended!
Best regards, John
womble
Posted 09/07/2009 - 13:50 Link
johnriley wrote:
This of course confirms that slide film has a narrow dynamic range, so if you can expose it properly you won't have a problem with digital.

Slides are a great learning tool and thoroughly recommended!

My first 35mm film ever was a roll of Kodachrome 64. RIP.
Kris Lockyear
It is an illusion that photos are made with the camera… they are made with the eye, heart and head. Henri Cartier-Bresson
Lots of film bodies, a couple of digital ones, too many lenses (mainly older glass) and a Horseman LE 5x4.

My website

Add Comment

To leave a comment - Log in to Pentax User or create a new account.