[DAZ]

tpeace.....I had the same experience. Basically, once you set the settings, you must then trigger the shutter the number of times you indicated (i.e. 5 times in your example). If you hit the menu it essentually wipes out the setting and assumes you did not want to take the multiple exposures. If it works correctly, you should be able to watch the counter take one shot away on your counter and then add it back to indicate it is writing to the previous image. Once you shoot the five shots, it then resumes normal "single exposure" shots. You must set this setting for each time you want to use the technique. At least that has been my experience so far.

A little additional info. When you are in multi mode there is a blinking symbol on the top LCD. If you wait to long to start taking the photo series it will go out of multi mode and the symbol go away. Once you start the series it will stay (how long I am not sure but a lot longer then it will wait for you to start the series) until you have taken all the photos in the series, assuming you don’t change one of the setting that takes it out of multi mode. If you do some thing to interrupt the series once you have started it will take however many you have taken (say 3 of the 5) and combine them and save the file.

DAZ

[DAZ]

There is a unified view on this.

Both, multi-exposure and bracketing, gather more light than a single exposure can do.

In the case of N exposures it is clear: you gather N x as much light.

In the case of 5x2EV bracketing, it is like 1 + 4 + 16 + 64 + 256 = 341 x as much light. For N<341, the bracketing method should win.

(The K-7 HDR does 3x3EV like 1 + 8 + 64 = 73 x.)

However, bracketing can lead to artifacts at contrast boundaries and the longer exposures may need a tripod. Therefore, multi-exposure can yield great HDR results (I use N=16 in such cases (just burst them away )).

So, in the first step, multi-exposure and HDR both yield high dynamic range pixel data. In the case of multi-exposure, by adding all pixels and dividing by N. In the case of bracketing, by cleverly choosing and normalizing the "best" pixels.

Now, the K20D/K-7 multi-exposure stops here and writes a final JPG by linearly(*) cutting the high dynamic range pixel data into 8 Bits.
(*) Well, JPG uses sRGB which compresses 3000 linear levels into 8 Bit...

The K-7 offers the second step in HDR processing as well: tone mapping.

The high dynamic range pixel data are non-lineearly mapped onto the 8Bit output space, by both compressing global contrast and expanding local contrast. The K-7 offers 2 different "strengths" to chose from.

Falconeye I am definitely not arguing with you just trying to see if I understand what is said here not whether or not someone will like the output.

When I take a photo in RAW the info is converted and stored in a 12 bit space. Because of noise not all of the DR of this 12 bits is available as the noise exceeds the signal before the least significant bit.

If I take multiple photos (like in the K20Ds multi mode) when the photos are combined (in the camera into RAW) signal is added but noise is not. Another way of saying noise is subtracted. This means that the photo can have more DR because more of the 12 bits range can be used (closer to the LSB). So overall the camera “looks” like it is using a lower ISO with less noise but I can’t exceed the 12 bit DR range. This would probably be more visible in photos that have more dark parts like night photos.

If I take photos (lets say 3) at different exposures and compress the information (tonal mapping) either in camera (K7) or PP into the 12 bit space of RAW (or 16 bit tiff) I can exceed the DR of 12 bits. This is kind of like of what happens when you go from RAW to JPG but with more DR total.

If I use something like D-range I am not changing the DR (I may even be lowering it) but applying a curve to bring up the shadows so that the eye can see more information. This may make more of the noise visible because there is less S/N at the lower levels.

Now a test for me to see if I get it. If I take a photo in multi mode and use D-range because of the apparent lower noise the output will have less shadow noise and more DR then if I had just taken 1 photo in D-range. Is this correct?

DAZ

[falconeye]

1. If I take multiple photos (like in the K20Ds multi mode) when the photos are combined (in the camera into RAW) signal is added but noise is not.

2. I can’t exceed the 12 bit DR range.

3. Now a test for me to see if I get it. If I take a photo in multi mode and use D-range because of the apparent lower noise the output will have less shadow noise and more DR then if I had just taken 1 photo in D-range. Is this correct?

You are asking detailed questions here.

1. "signal is added but noise is not": The correct statement would read: signal is added linearly but noise is added by its square root only. If you combine images, noise increases as well, but slowlier than the signal.

2. Most people think this way and it doesn't really matter for this discussion. But it isn't true actually. Think of a dice and you record only 1 Bit "hit"/"miss": by taking the statistics, you will see a 1/6 distribution for hit and have resolved almost 3 Bits from 1 Bit measurements. Therefore, 12 Bits don't impose any kind of limit here.

3. You are correct if the following is true: The K20D/K-7 adds the N images in RAW space rather than JPG space and it takes individual images at an exposure normal for a single shot, i.e., it does not underexpose by 1/N. I believe this to be true but I am not 100% sure. I didn't check because personally, I prefer to burst N individual images and combine in PhotoAcute which saves me the tripod. 16 night shots at ISO1600 each give amazing results

[DAZ]

Thanks for answering my questions falconeye.

You are asking detailed questions here.

I was trying to ask specific questions so as to not waste your time.

1. "signal is added but noise is not": The correct statement would read: signal is added linearly but noise is added by its square root only. If you combine images, noise increases as well, but slowlier than the signal.

Slap head! Should have seen that one. The noise is not just going to go away. Yes it increases but not much as the signal but I did not know the noise only increased by its square root.

2. Most people think this way and it doesn't really matter for this discussion. But it isn't true actually. Think of a dice and you record only 1 Bit "hit"/"miss": by taking the statistics, you will see a 1/6 distribution for hit and have resolved almost 3 Bits from 1 Bit measurements. Therefore, 12 Bits don't impose any kind of limit here.

I had to think about this one for awhile. If I understand correctly you are saying from the point of view of DR only, 12 bits is not a limit as it is wider then is needed.

3. You are correct if the following is true: The K20D/K-7 adds the N images in RAW space rather than JPG space and it takes individual images at an exposure normal for a single shot, i.e., it does not underexpose by 1/N. I believe this to be true but I am not 100% sure. I didn't check because personally, I prefer to burst N individual images and combine in PhotoAcute which saves me the tripod. 16 night shots at ISO1600 each give amazing results

I know the K20D can save the data in RAW for multi mode exposures and I believe the K7 saves its multi exposure HDR in RAW also. The example was only to see if I understood what is happening here. So if I had to choose my self I would make sure I had exposed as best I could then combine on the computer. The computer has much more power to use and give more control.

Now I have to practice some of this so I can determine when to use it best.

Again thanks for the answers.

DAZ

[alman]

I read on here somewhere that by using multi exposure on the K 20 you can acheive the same effect as using an ND filter with long exposure times. An example of this is to get that silky water look. I have tried this technique out myself and was very impressed with the results using 9 exposures. Anything moving (water, clouds) had that dreamy look and anything static (rocks etc) were as sharp as a tack. Great effect! Now there is a question at the end of all this....how does this feature differ from the 3 shot HDR on the K7?
(BTW I explained this technique to a local Canon dealer and he couldn't believe it...never heard of a camera that could do that he said...chalk another one up for Penatax and the fab K20!)

The k-7 HDR feature does the same thing without the need of further processing with a computer. It gives better results sometimes than multiple shots with computer processing and sometimes it doesn't. I use it frequently and like it.

[falconeye]

I had to think about this one for awhile. If I understand correctly you are saying from the point of view of DR only, 12 bits is not a limit as it is wider then is needed.

Since you have thought about it for a while.

What I meant was that you would be able to achieve, e.g., 14 Bit dynamic range from multiple images each having 12 Bit with 11 Bit dynamic range only (and noise in the least significant 12th bit). You would need 4^3 or 64 images to gain these extra 3 Bits and there would be no 12 Bit barrier.

Most people cannot understand this. They cannot accept that 64 images with 12 Bit depth can yield anything but another image with 12 Bit depth. A proper understanding that noise here is actually helping is crucial. Because indeed, 64 identical images would not yield any more information than a single one.

In practice though, you will hit the barrier of pattern noise much earlier than these high dynamic range levels. This limits the number of usable images to about 16 with a K20D.

[Trainman Dave]

K10D can do it, too! But I do keep thinking about some of the new features in the 20D and 7...

[DAZ]

Since you have thought about it for a while.

What I meant was that you would be able to achieve, e.g., 14 Bit dynamic range from multiple images each having 12 Bit with 11 Bit dynamic range only (and noise in the least significant 12th bit). You would need 4^3 or 64 images to gain these extra 3 Bits and there would be no 12 Bit barrier.

Most people cannot understand this. They cannot accept that 64 images with 12 Bit depth can yield anything but another image with 12 Bit depth. A proper understanding that noise here is actually helping is crucial. Because indeed, 64 identical images would not yield any more information than a single one.

In practice though, you will hit the barrier of pattern noise much earlier than these high dynamic range levels. This limits the number of usable images to about 16 with a K20D.

I think I can get that. It is kind of like the way in forward error correction you can correct for more errors then FEC bits you sent. I don’t totally under stand the math but I have seen it work enough to just push the “I believe button”.

As you say pattern noise is the limiter here (I assume you got that number by experimentation as pattern noise is camera type dependent) is 16 the limit just because of noise or just the point of diminishing returns. 16 is a large number of extremely big file to work with. My little system starts to gag at 4-8 files of that size.

On a side note. Do I have to stack all the files simultaneously or can I group them. For example, can I stack then in groups of 4 and then stack the 4 groups? This is of course using 16 unique files but I would only need to deal 4 at a time. Or for that matter have the camera do a group of 8 and then a second group of 8 then I just have to combine the 2 files in the computer.

Thanks

DAZ

[falconeye]

With about 16 images taken at ISO1600 and a K20D, pattern noise starts to disturb me, esp. the magenta glow. The following image is a test example of a dark street freehand night shot pushed to ISO 25600 (16 x ISO 1600 and +4EV in LR). No noise but disturbing glow. The dynamic range in the center is about 13 EV. The recent work of GordonBGood tries to fix it though.

[IMGWIDELEFT]http://www.falklumo.com/downloads/pub/2009.06/IMGP6119-34.jpg[/IMGWIDELEFT]

Note that the above image is with exposure pushed up +4EV in post processing!

Do I have to stack all the files simultaneously or can I group them.

I use the PhotoAcute software to stack my images. It is fast enough and does the alignment for freehand shots as well.

As for your question: Stacking 4 times 4 images, then stacking the 4 results, is equivalent to stacking the 16 images in the first place. Provided you do it in 16 Bit.

[insulinguy]

went out to try this today and i have to say - something isnt right. i ended up with 9 of the exact same image rather than 9 images merged into one...i did the auto EV as well - to no avail. i thought you snap the images then the camera produced ONE image. am i missing something here? i use the K20

[DAZ]

With about 16 images taken at ISO1600 and a K20D, pattern noise starts to disturb me, esp. the magenta glow. The following image is a test example of a dark street freehand night shot pushed to ISO 25600 (16 x ISO 1600 and +4EV in LR). No noise but disturbing glow. The dynamic range in the center is about 13 EV. The recent work of GordonBGood tries to fix it though.

[IMGWIDELEFT]http://www.falklumo.com/downloads/pub/2009.06/IMGP6119-34.jpg[/IMGWIDELEFT]

Note that the above image is with exposure pushed up +4EV in post processing!


I use the PhotoAcute software to stack my images. It is fast enough and does the alignment for freehand shots as well.

As for your question: Stacking 4 times 4 images, then stacking the 4 results, is equivalent to stacking the 16 images in the first place. Provided you do it in 16 Bit.

Thank for all the help. That is all the questions I have. I going to look into that PhotoAcute.

DAZ

[DAZ]

went out to try this today and i have to say - something isnt right. i ended up with 9 of the exact same image rather than 9 images merged into one...i did the auto EV as well - to no avail. i thought you snap the images then the camera produced ONE image. am i missing something here? i use the K20

Yes I am afraid you did something not correct. It is easier to start from the beginning.. From the top.
Push MENU.
Arrow down to Multi-exposure. It should say Off.
Right arrow one time. The display will have 2 lines. First Number of shots it will say off. Second line Auto EV Adjust with a check box.
Arrow down one to the Auto EV Adjust line. This is how you turn on the Auto EV so you don’t have to do the math on what exposure you need to set. If you arrow over right 1 it will put a check in the box so do that.
Now up arrow to the number of shots line.
Now if you right arrow 1 time a drop down menu will appear.
Arrow down to the number of time you want to shoot. Then push OK
Now you have to be a little careful from here as you can turn off the Mulri-exposure if you push the incorrect button. Push MENU 2 time.
On the top LCD will be a new symbol above the drive mode and to the left of the aperture number.

If you let the camera go the sleep you will lose the setting. If you go back into the MENU to look at it you will lose the setting.

When you take the first of the photos the camera will display (if you have it set up to display the photo on the back LCD) on the back LCD a photo as if you were not doing multi. This is the exposure 1 with the settings so you can see if you like the aperture, shutter speed and ISO for overall exposure. The symbol on the top LCD will also be blinking.

When you take the next photo the number of remaining shot on the card (not how many left in the multi group) will go down then back. The back display will flash the combined of the 2 exposures (so if it is more then 2 it will look under exposured). As you take more the camera will build them up.

Keep shooting until you get to the number you set. The only way to know how many to go is count. When you get to the last photo the camera combine the last and display it like the others after it saves it to the card. The multi symbol will go off.

If you do something that takes the camera out of multi before you get to the last number you set the camera will recompute and save what you have so far.

DAZ

[Class A]

You would need 4^3 or 64 images to gain these extra 3 Bits...

Is it 4^3 rather than 2^3 because the noise only grows with the square of the number of images and you are interested in increasing the bits above the noise?

A proper understanding that noise here is actually helping is crucial.

Quite similar to "dithering" (intentional adding of noise in order to reduce quantization errors) in digital audio.

[insulinguy]

i was using the remote with the multi exposure so i am sure one must have off set the other - ill try it again. thanks for the walk thru - reads much better than the manual! does anyone know if you can do multiple images in the same frame? like sequencing? i think the D90 does it but ive always been a pentaxian so even though it was a feature i wanted - it wasnt enough to get me to switch to nikon.

[thomascdk]

I'm pretty sure you have to set it to remote before setting it to multiple exposure...