Pages: [1]   Go Down

Author Topic: reducing highlight clipping  (Read 2663 times)

bwana

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 309
reducing highlight clipping
« on: May 14, 2014, 12:14:07 pm »


Is there a process/technique that allows compressing a broad dynamic range image onto a sensor with smooth tonal gradations.

One way this could be accomplished would be if the camera software could generate a luminosity mask. As Tony describes here . such a 'self-feathering mask' would act as grad filter that is custom shaped to the highlights of the image.

IS anyone aware of a camera/hacked firmware that can do this?
or perhaps someone has solved this problem another way? Taking the SD card out of the camera, making a luminosity mask of the image, saving that mask as another image to camera sd card, then doing a multiple exposure onto that image of the same scene?
Logged

EricV

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 270
Re: reducing highlight clipping
« Reply #1 on: May 14, 2014, 01:36:42 pm »

Highlight clipping means the exposure delivers more photons to the sensor than the pixels can hold.  In normal digital sensors, all pixels across the sensor have the same photon capacity, and this capacity cannot be changed by software.  So your software luminosity mask idea will not work.

A mask which blocks some of the excess photons from reaching the sensor would solve the problem.  A graduated neutral density filter is one common hardware solution which does this.  But it is not very programmable.

Another idea is to reduce the exposure to the point where highlight clipping does not occur, then do something else to recover the shadows which are now too dark.  Multiple exposures and high dynamic range (HDR) post-processing is a common solution along these lines.

In keeping with your desired in-camera software solution, you want a sensor with programmable sensitivity (ISO setting) at the pixel level.  Then you could keep the exposure below the clipping level for normal pixels, and raise the electronic gain of pixels in the dark areas.  A pre-exposure could determine which pixels need low gain and which pixels need high gain.  No digitial sensor has this capability, and it would greatly complicate the electronics to create such a sensor. 

A few cameras do contain a mix of pixels with different photon capacity, but the pattern is not programmable.  The idea is to expose so that the low capacity pixels capture the shadows, then reconstruct the clipped highlights from the high capacity pixels.
Logged

SZRitter

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 384
Re: reducing highlight clipping
« Reply #2 on: May 14, 2014, 01:59:29 pm »

If I remember correctly, Olympus has a patent similar to what you want.

Since they are already doing sensor readings mid-exposure, it actually wouldn't be too hard. You could mark off when the pixel maxed out, and with some funky math that is way beyond my skill level, I am sure you could get it.
Logged

SZRitter

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 384
Re: reducing highlight clipping
« Reply #3 on: May 14, 2014, 02:23:56 pm »

Logged

bwana

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 309
Re: reducing highlight clipping
« Reply #4 on: May 14, 2014, 03:03:32 pm »

thank you for your replies to my weird question. the problem i find with hdr is that it is too coarse. when several exposures are made at different EVs, the histogram is shifted by significant amounts. So when you combine these images, the histograms are summed. A problem arises because the smooth roll off from light to dark in any given image now becomes 'bumpy' with numerous peaks - each little peak represents a peak from each different exposure. This tonal 'bumpiness' gives the image a weird appearrance with a fluctuating luminance instead of a smooth gradation.; To eliminate this bumpiness and 'tonal confusion' where tones are crossing ansel adams' zones, many, many exposures need to be summed-not a practical solution.

I am waiting for the new olympus sensor.
Logged

Jim Kasson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2370
    • The Last Word
Re: reducing highlight clipping
« Reply #5 on: May 14, 2014, 03:22:10 pm »

thank you for your replies to my weird question. the problem i find with hdr is that it is too coarse. when several exposures are made at different EVs, the histogram is shifted by significant amounts. So when you combine these images, the histograms are summed. A problem arises because the smooth roll off from light to dark in any given image now becomes 'bumpy' with numerous peaks - each little peak represents a peak from each different exposure. This tonal 'bumpiness' gives the image a weird appearrance with a fluctuating luminance instead of a smooth gradation.; To eliminate this bumpiness and 'tonal confusion' where tones are crossing ansel adams' zones, many, many exposures need to be summed-not a practical solution.

If you underexpose and use the Exposure control in Lightroom PV2012 to brighten the image, it will compress highllights much the way film does. This works especially well with "ISOless" cameras like the a7, a7R, D800, D4, and M240.

http://blog.kasson.com/?p=2960

Jim

EricV

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 270
Re: reducing highlight clipping
« Reply #6 on: May 14, 2014, 05:06:44 pm »

A problem arises because the smooth roll off from light to dark in any given image now becomes 'bumpy' with numerous peaks - each little peak represents a peak from each different exposure.
That is not how proper HDR processing works.  The software should first combine the images, with proper exposure scaling, to extend the dynamic range without introducing any bumps.  Tone mapping then compresses the dynamic range back to something which can be displayed or printed, again without introducing any bumps.  Here is a nice reference:
http://www.cambridgeincolour.com/tutorials/high-dynamic-range.htm
Logged

EduPerez

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 700
    • Edu Pérez
Re: reducing highlight clipping
« Reply #7 on: May 15, 2014, 06:37:04 am »

If you underexpose and use the Exposure control in Lightroom PV2012 to brighten the image, it will compress highllights much the way film does. This works especially well with "ISOless" cameras like the a7, a7R, D800, D4, and M240.

http://blog.kasson.com/?p=2960

Jim

Canon has also incorporated a similar technique to its cameras, and calls it "highlight tone priority".
Logged

bwana

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 309
Re: reducing highlight clipping
« Reply #8 on: May 15, 2014, 09:39:23 am »

That is not how proper HDR processing works.  The software should first combine the images, with proper exposure scaling, to extend the dynamic range without introducing any bumps.  Tone mapping then compresses the dynamic range back to something which can be displayed or printed, again without introducing any bumps.  Here is a nice reference:
http://www.cambridgeincolour.com/tutorials/high-dynamic-range.htm

this article explains my findings better
http://blog.mingthein.com/2014/04/14/hdr-zone-system-dynamic-range/

the bottom line is this: in a colorspace/display device where the rendering options are a subset of the capture options, you have to compress tones if you want to include the whole range. evenly distributed compression is fruitless because you cannot display everything. tone mapping (unequal tonal compression based on brightness levels) results in areas of low contrast and high contrast distributed weirdly throughout the image. if i wanted to do hdr my way, i would do tone mapping based on parts of the picture- some areas of my photo would benefit from different tone mapping than other areas, even though they might have similar tone ranges. do you know of any program that might allow me to 'paint' in a certain kind of tone mapping in one part of the image and a different kind of tone mapping in another? yes i know i can make different hdr images and combine them in layers in PS and mask, but i just do not have the patience and the time to experiment so slowly. I am more of a 'dodge and burn' kind of guy. it would be nice if i could assign different tone mapping algorithms to my brush as well.
Logged

Lightsmith

  • Full Member
  • ***
  • Offline Offline
  • Posts: 197
Re: reducing highlight clipping
« Reply #9 on: May 18, 2014, 01:46:22 am »

It is not accurate to talk about combining "histograms" to create a image with greater dynamic range or less clipping. There is only so much information available in terms of reflected light for any given scene. A shot that uses an exposure for the shadows may result in loss of data at the opposite end of the scale but that is easily overcome with a second shot that is exposed for the highlights.

There is greater dynamic range available than many people realize. Where mistakes are made is in underexposing for fear of clipping highlights and relying 100% on the histogram to verify this. Shooting RAW the histogram can make it appear that data has been lost for the highlights when this is actually not the case. Going out in bright sunlight and shooting three shots at -1.0 EV and 0.0 EV and +1.0 EV in RAW capture and then opening them with a RAW converter you will find that there is considerably more available information in the +1.0 EV than the -1.0 EV and in some instances the +1.0 EV shot will have more information than the 0.0 EV shot which is actually an underexposed shot.
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: reducing highlight clipping
« Reply #10 on: May 18, 2014, 02:32:07 am »

Depends a lot on your camera. Using a tool like Rawdigger shows correct histograms.

I would recommend a small article I have written some time ago:
http://echophoto.dnsalias.net/ekr/index.php/photoarticles/63-lot-of-info-in-a-digital-image

Best regards
Erik

It is not accurate to talk about combining "histograms" to create a image with greater dynamic range or less clipping. There is only so much information available in terms of reflected light for any given scene. A shot that uses an exposure for the shadows may result in loss of data at the opposite end of the scale but that is easily overcome with a second shot that is exposed for the highlights.

There is greater dynamic range available than many people realize. Where mistakes are made is in underexposing for fear of clipping highlights and relying 100% on the histogram to verify this. Shooting RAW the histogram can make it appear that data has been lost for the highlights when this is actually not the case. Going out in bright sunlight and shooting three shots at -1.0 EV and 0.0 EV and +1.0 EV in RAW capture and then opening them with a RAW converter you will find that there is considerably more available information in the +1.0 EV than the -1.0 EV and in some instances the +1.0 EV shot will have more information than the 0.0 EV shot which is actually an underexposed shot.
Logged
Erik Kaffehr
 

Telecaster

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3686
Re: reducing highlight clipping
« Reply #11 on: May 18, 2014, 04:34:47 pm »

With newer cameras I rarely find "not enough DR" to be an issue. I'm typically throwing away tonal data rather than wishing I had more of it. I do like that Leica's M8(.2), which I've been using a lot lately, derives playback histograms from the RAW data. Often an exposure will have blown highlight areas according to the immediate post-shot display, but these will disappear or at least lessen in the RAW-derived display. As the M8 has enough DR for my needs, but not an excess of it, optimal exposure is more important to me with this camera. But with Sony's A7r, for example, I see no real-world gain in shooting a more accurate ETTR as opposed to just going with the camera's histogram.

When DR is an issue I like to establish a base exposure just short of clipping (according to the camera), then also shoot a +1 & +2. Maybe all the way up to +4 depending on what the shadow end of the histograms looks like. Then I'll typically blend the best + exposure with the base one. I started doing this with scanned transparencies back in the mid 1990s and find it still usually works fine with purely electronic data.

-Dave-
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: reducing highlight clipping
« Reply #12 on: May 19, 2014, 02:40:41 am »

Hi,

I would say that DR has seldom been a problem for me, like half a dozen times in 70000 exposures. Now when I am also shooting P45+ it seems a bit more limited.

In general I feel that having a broad DR is an advantage.

Regarding highlight clipping, my feeling is don't clip. I don't feel reconstructed clouds and similar look any good.

Best regards
Erik


With newer cameras I rarely find "not enough DR" to be an issue. I'm typically throwing away tonal data rather than wishing I had more of it. I do like that Leica's M8(.2), which I've been using a lot lately, derives playback histograms from the RAW data. Often an exposure will have blown highlight areas according to the immediate post-shot display, but these will disappear or at least lessen in the RAW-derived display. As the M8 has enough DR for my needs, but not an excess of it, optimal exposure is more important to me with this camera. But with Sony's A7r, for example, I see no real-world gain in shooting a more accurate ETTR as opposed to just going with the camera's histogram.

When DR is an issue I like to establish a base exposure just short of clipping (according to the camera), then also shoot a +1 & +2. Maybe all the way up to +4 depending on what the shadow end of the histograms looks like. Then I'll typically blend the best + exposure with the base one. I started doing this with scanned transparencies back in the mid 1990s and find it still usually works fine with purely electronic data.

-Dave-
Logged
Erik Kaffehr
 

Ajoy Roy

  • Full Member
  • ***
  • Offline Offline
  • Posts: 117
Re: reducing highlight clipping
« Reply #13 on: May 26, 2014, 09:38:22 am »

Unlike analog media, in digital once the sensel is saturated it can no longer collect photons. That is highlight clipping. So you always have be just below the clipping level.

The current Nikon (and Sony) cameras which use Sony sensors have a DR of 13-14EV, compared to 8-10 EV of older cameras. So in effect the sensor is capturing data which would have required HDR in older cameras.

I have Nikon D3300, and the Nikon exposure give me at least 1EV of headroom, so I can recover upto 1EV in the RAW file. Couple that with at least 13EV of DR, in most situations you do not need HDR as you can recover deep shadows. The DF and D4 have even better DR so you have data for 14+EV with you. The trick is to expose for the brightest area and then recover the shadows.
Logged
Ajoy Roy, image processing

PeterAit

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4559
    • Peter Aitken Photographs
Re: reducing highlight clipping
« Reply #14 on: May 26, 2014, 12:40:49 pm »

thank you for your replies to my weird question. the problem i find with hdr is that it is too coarse. when several exposures are made at different EVs, the histogram is shifted by significant amounts. So when you combine these images, the histograms are summed. A problem arises because the smooth roll off from light to dark in any given image now becomes 'bumpy' with numerous peaks - each little peak represents a peak from each different exposure. This tonal 'bumpiness' gives the image a weird appearrance with a fluctuating luminance instead of a smooth gradation.; To eliminate this bumpiness and 'tonal confusion' where tones are crossing ansel adams' zones, many, many exposures need to be summed-not a practical solution.


I am curious as to how you are doing the HDR that you describe as unsatisfactory. I believe that HDR blending algorithms have gotten way beyond the point of simply summing (or averaging) the histograms. Can you post a sample image with the bumpiness?
Logged
Pages: [1]   Go Up