Skip navigation
This discussion is archived
Currently Being Moderated

Canon MKIII Highlight tone recovery V ACR recovery

Oct 14, 2007 2:24 AM

I can see the point of using the 1D MKIII's HTR feature with JPEGS for high contrast range subjects to hold important highlight detail.

What is the consensus re in-camera HTR for RAW, versus no in-camera HTR and using highlight recovery in AR V4 on RAW files? (less chance of shadow noise with ACR as highlight specific adjustment)

Are Canon merely reducing the exposure at the highlight end and increasing overall exposure from shadows up the scale hence their advisory notice that HTR technology may increase shadow noise?
 
Replies
  • Currently Being Moderated
    Oct 14, 2007 5:44 AM   in reply to (Nick_Walker)
    In camera, all this is happening in raw mode with HTP is the exposure is reduced by 1-fstop.

    This creates a raw file with 1 extra stop of headroom. Exactly how the raw converter handles this is up to the raw converter. Camera Raw *by default* boosts the exposure slider zero point back up by one stop.

    For many images, it is absolutely the correct thing to do. If you are taking photographs of normal scene, and have the camera set in HTP mode, then this is EXACTLY what you want. Rolling off the highlights like DPP does would just create a muddy image.

    If you have an image that you needs rolled off highlights rather than boosted highlights, that is trival to do Camera Raw. Just boost the recovery slider (for versions of Camera Raw with that control). Or you can do a combination of reducing the exposure slider and increasing the brightness slider, which has the same effect.
     
    |
    Mark as:
  • Currently Being Moderated
    Oct 19, 2007 3:51 AM   in reply to (Nick_Walker)
    HTP is the same as "expose to the left". It increases highlight headroom by reducing exposure by 1 stop. The downside is higher noise.

    For raw files, you can get the same effect in any camera by dialing in -1 stop of exposure compensation and adjusting to compensation in the raw conveter.
     
    |
    Mark as:
  • Currently Being Moderated
    Oct 19, 2007 8:29 AM   in reply to (Nick_Walker)
    >For raw files, you can get the same effect in any camera by dialing in -1 stop of exposure compensation and adjusting to compensation in the raw conveter.

    True enough, but the camera preview will be dark and the histogram shifted to the left. Nick, how does the 1DMIII handle this situation? I would presume that the preview would be brightened by 1 f/stop, but what about the histogram? Personally, I would prefer to have the histogram show the true status of the exposure and not some shifted value.

    Bill
     
    |
    Mark as:
  • Currently Being Moderated
    Oct 19, 2007 9:02 AM   in reply to (Nick_Walker)
    > As HTP only routinely underexposes by 1 stop, and there is no advanced sensor technology taking place, it appears to me that slight over exposure followed by accurate highlight recovery in ACR V4 would be preferrential as there would be slightly less noise at he shadow end

    The exposure with HTP On is the same as with HTP Off, but the ISO will be halved with HTP On (that's the reason for HTP working only from ISO 200 up). The loss in the raw data is the least significant bit. As the 1DMkIII creates 14-bit raw data, this loss is practically meaningless (assuming that the extra bits or at least the first of them is not random).

    On the other hand, HTP is important rather when recording JPEG in-camera. When recording raw data, DPP applies a "highlight recovery" curve to HTP images, but this can be done w/o HTP as well. One could even argue that this is a disadvantage, because one can not turn off this function of DPP, unlike ACR's "recovery".

    Memory cards are dirt cheap, the camera is fast. IMO the best way is to shoot with exposure bracketing and select the most fitting one in raw processing.

    > Personally, I would prefer to have the histogram show the true status of the exposure and not some shifted value

    What is "true status"? The in-camera histogram is based on the camera settings, which have no impact on the raw data. I would prefer raw histogram shown in-camera, at least as an option, but I don't think many customers share this preference.
     
    |
    Mark as:
  • Currently Being Moderated
    Oct 19, 2007 9:20 AM   in reply to (Nick_Walker)
    >Personally, I would prefer to have the histogram show the true status of the exposure and not some shifted value

    >What is "true status"? The in-camera histogram is based on the camera settings, which have no impact on the raw data. I would prefer raw histogram shown in-camera, at least as an option, but I don't think many customers share this preference.

    This issue has been covered many times. Since the histogram is derived from the JPEG preview with most cameras, you should set the contrast to low. Contrast affects the quarter tones much more than the extremes of the histogram. If you want to see the channels without white balance, at least with Nikon cameras, you can upload a special white balance to the camera where the red and blue multiplers are set to unity (UniWB). Also you can upload a TRC to the camera to undo the gamma encoding, but I've never tried this.

    If you do a bit of experimenting, you can correlate the camera histogram with the raw file and have a pretty good idea of what is going on with the raw file, and you can gain quite a bit of control over the histogram. Some cameras indicate clipping in the histogram when there is none in the raw file, and this can be corrected with the proper TRC.

    IMHO, a raw linear histogram is not a good idea, since the data would be scrunched up on the left. A log base 2 histogram would correspond to f/stops and that would be my preference and also corresponds with human luminance perception, which is log.

    >The exposure with HTP On is the same as with HTP Off, but the ISO will be halved with HTP On (that's the reason for HTP working only from ISO 200 up). The loss in the raw data is the least significant bit. As the 1DMkIII creates 14-bit raw data, this loss is practically meaningless (assuming that the extra bits or at least the first of them is not random).

    If you halve the exposure, shot noise will be increased by a factor of 1.4 across the tonal range. Usually dynamic range is limited by the noise floor, rather than by quantization. The 14 bit ADC wouldn't address the increased shot noise, but it might reduce the read noise.
     
    |
    Mark as:
  • Currently Being Moderated
    Oct 19, 2007 9:41 AM   in reply to (Nick_Walker)
    >Nick, how does the 1DMIII handle this situation? I would presume that the preview would be brightened by 1 f/stop, but what about the histogram?

    The in-camera histogram is usually based on the preview file (jpg).
     
    |
    Mark as:
  • Currently Being Moderated
    Oct 19, 2007 12:09 PM   in reply to (Nick_Walker)
    > If you want to see the channels without white balance, at least with Nikon cameras, you can upload a special white balance to the camera where the red and blue multiplers are set to unity (UniWB). Also you can upload a TRC to the camera to undo the gamma encoding, but I've never tried this

    These are great options; unfortunately, Canon does not offer such. With my next camera I will spend some time to figure out, based on test shots, which temperature and bias setting comes the closest to a "leave it" action.

    > If you halve the exposure

    but you don't, at least not by turning on HTP.
     
    |
    Mark as:
  • Currently Being Moderated
    Oct 19, 2007 1:31 PM   in reply to (Nick_Walker)
    >If you halve the exposure

    >>but you don't, at least not by turning on HTP.

    Yes, in re-reading your description I see that HTP uses the same exposure but decreases the amplifier gain by one stop so as to give extra headroom. This is why the method is not offered at base ISO. In view of this, I think that Thomas Knoll's explanation is not quite correct. Noise should not increase.
     
    |
    Mark as:
  • Currently Being Moderated
    Oct 19, 2007 2:35 PM   in reply to (Nick_Walker)
    > I think that Thomas Knoll's explanation is not quite correct

    From the point of ACR it's the same: in order to compensate for it, you do the same as if the exposure had been lower.

    > Noise should not increase

    That's right, in the sense of the sensor. However, when the exposure will be increased by one stop, the raw values get doubled in effect. Somewhat simplified, the last bit of the raw values becomes generally zero. The effect can appear as noise.

    On the other hand, if the 13th bit of the Canon 1DMkIII is not random, that is a substitute for the "lost" bit (this question has not been analyzed in detail yet).

    On the third hand, if you look at the meaning of the 12th bit, you find that it is virtually meaningless, unless you are doing extensive adjustments to the raw data. So, the effect may or may not be noticable.
     
    |
    Mark as:
  • Currently Being Moderated
    Oct 20, 2007 9:31 AM   in reply to (Nick_Walker)
    >From the point of ACR it's the same: in order to compensate for it, you do the same as if the exposure had been lower.

    That's right, but as I understand HTR, and correct me if I'm wrong, to duplicate the effect in any camera, you would have to decrease the ISO by half and dial in -1 f/stop exposure compensation. This would keep the exposure the same, but decrease the amplifier gain so that the ADC is operating at only half scale, giving an extra stop of headroom. The preview would be dark and the histogram would show the increased headroom.
     
    |
    Mark as:
  • Currently Being Moderated
    Oct 20, 2007 11:41 AM   in reply to (Nick_Walker)
    > to duplicate the effect in any camera, you would have to decrease the ISO by half and dial in -1 f/stop exposure compensation

    Right, that's exactly the effect of HTP. What I meant above was, that when you are processing the image in ACR, you can't select between compensating for lower ISO and compensating for lower exposure.

    So, even though not the exposure gets reduced by HTP, the action in ACR is increasing the exposure (which too is technically incorrect...).

    > The preview would be dark and the histogram would show the increased headroom

    and the JPEG output would be quite useless. While HTP on the raw level means only convenience, the real advantage is for JPEG shooters.
     
    |
    Mark as:
  • Currently Being Moderated
    Dec 5, 2007 5:55 PM   in reply to (Nick_Walker)
    Bill Janes - 1:31pm Oct 19, 07 PST (#11 of 14)
    ..................
    Yes, in re-reading your description I see that HTP uses the same exposure but decreases the amplifier gain by one stop so as to give extra headroom. This is why the method is not offered at base ISO. In view of this, I think that Thomas Knoll's explanation is not quite correct. Noise should not increase.
    ______________________

    Hi Bill - actually it does increase. I've been testing my new 1DsMKIII for a couple of days now. Yesterday I had HTP enabled and at ISO 200 I found the 3/4 tones much noisier in both the raw and processed raw images (using CR4.3) than I would have expected based on the other type of noise being made about this camera. So I consulted the manual (FWIW) and sure enough Canon warns us that enabling this feature will cause more noise. So this morning I returned to the same spot(a group of tall blackish bank buildings with security guards who don't like people making photographs there) and made another set of images with HTP disabled. I procsssed a number of them this evening and there is definitely MUCH less noise - in fact barely any at ISO 200. I also measured the noise differences using Noise Ninja auto-profiling. Whereas yesterday's measurements were in the range of 17 to 23, today's are in the range of 9 to 14, in both cases about 3/4 of it being Luminosity Noise, 1/4 Color. In CR 4.3 I have noise set at the defaults - Color at 25% and Lum 0 (to preserve maximum image detail).

    Cheers,

    Mark
     
    |
    Mark as:

More Like This

  • Retrieving data ...

Bookmarked By (0)