On the dynamic range of RAW...

Started Nov 30, 2004 | Discussions thread
David SL Veteran Member • Posts: 3,474
Re: COLOR fares not as good

Eric_17 wrote:

I didn't read the whole thread here, but I don't see any mention of
anyone exploring the effects of this process with color.

Increasing the dynamic range (decreasing contrast) inherently
desaturates color and adds more grey to an image. (And by the same
principle, decreasing dynamic range (increasing contrast)
inherently increases color saturation.)
It seems to me that
Brian's results are falsely spectacular (don't get me wrong, + -4
EV really is impressive), but if he had tried this with a color
image, the results would not have been the magazine-worthiness the
black & white one is.

Not necessarily, keep in mind that the expose and blend technique does not "increase" the dynamic range of the camera. Instead the amount of tones that can be represented in the final image is closer to what the eye saw than what the camera was able to envision in any one exposure. The camera's overall dynamic range specifies how many tones of color it is able to span from darkest to lightest. The more distinct tones that can be applied to pixels from darkest dark to brightest bright area, the highier the dynamic range. DR tells you not the number of colors you have in total (that's what the color space gamut/file gamut tells you) but rather the number of distinct tones from that palette from which you can choose for any given exposure. In the case of a jpeg image, the total palette size is just 16 million colors, the number of colors that can actually be used from that palette depends on the cameras dynamic range.

This is a very important and subtlety as it allows us to answer the question definitively. Does increasing DR desaturate an image? The answer is it depends on how you are increasing the DR , over what domain. If you are taking a single exposure and by curve manipulations pulling up shadows and midtones to match well exposed highlights and then blending the adjusted exposure(s) with the principle one, then you will be reducing contrast toward the shadows and midtones. This occurs because of a high level of quantization error associated with pulling up shadow and midtone detail, in extreme cases you'll notice this error shows up as noise in the pulled up shadows. A RAW image is effected in this way by a smaller degree, it has a highier tolerance for post processing adjustments as it can render more distinct tones per pixel location . (10, 12bit versus 8bit per pixel in JPEG) The overall palette is still(assume same color space for each shot) the same size but the RAW uses finer shades to "paint" the scene detail.

With this explanation in mind, recall that Brian did not use a single exposure, he used multiple exposures. Let's make a hypothetical example, that the D70 is able to provide 10 million tones of color from the available palette of 16 million(assuming a jpeg shot). Brian's highlight exposed version of the shot consisted of many bright valued tones to render the detail into the highlight, as the tone slots began filling up to 10 million, the camera had to be more sparse with how the tones were used into the shadows (since that's how the exposure was set up). This leaves the midtones and shadows with few available tones to represent them adequately, in some cases this can result in zeroed out pixels. Now, for the second picture, importance is given to the shadows. In this case the cameras 10 million tone slots are filled by mostly dark tones which fill up toward the bright end forcing the opposite problem, sudden transition to very bright areas (255) in the highlights. By mixing the two exposures, Brian increases the final DR of the image without desaturating the image, since both portions of the mixed image are properly exposed. He took the optimal tones for each portion of the image from the separate exposures and combined them, no manipulation of the tones was done beyond that and no quantization error results since no tone or curve adjustments are needed. You can confirm this readily by looking at the example images provided at luminous landscapes and elsewhere, the images are normally perfectly saturated and well exposed from dark to light. When done right merged images using multiple accurate exposures for shadow,midtone, highlights are realistic in a sense we never see without using our own eyes.


-- hide signature --

Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow