Red Flowers Photography: Real Problems or Deceptive Previews?

Red Flowers Preview: Looking at the Funhouse Mirror; or What You See is Not What You Got

Shooting red flowers... Always a problem... Let's start with...

Now, back to the beginning...

We started by shooting a series of RAW+JPEG pairs of a red flower, varying the shutter speed.

While Taking Photos of Flowers...

Let's take a quick look at the in-camera histograms:

In-camera histograms for shots from 4625 (top left) to 4630 (bottom right)

If we were to base our decision as to which shots to keep on the in-camera histogram, we, of course, would keep shot 4625 (the histogram at the top left corner) as almost-perfectly exposed to the right, and immediately discard shot 4630 (the histogram at the bottom right corner) as heavily clipped in the red channel.

But let's examine this presumably "heavily overexposed in red" shot, number 4630; we will see that clipping on the in-camera histogram is not the same as genuine overexposure of RAW.

While Culling/Selecting Photos of Flowers...

... Using Out-of-Camera JPEGs and the JPEG Histogram

What if we were to use the, unfortunately, rather traditional way of selecting RAW photos based on their JPEG previews and the JPEG histogram?

Shot 4630, external JPEG opened in Adobe Photoshop

You can see exactly what was described in the blockquotes - a kind of red blob instead of a red flower, with the channel R histogram climbing the wall all the way to the top (exactly as the in-camera histogram does, because the in-camera histogram is based on the out-of-camera JPEG, not RAW).

• If you would like to know in more detail why you shouldn't rely on JPEG previews and JPEG histograms for selecting ANY photos (and not only of red flowers), please have a look at our articles:
• "The Three Most Obvious Reasons to Look at RAW and Not Cull Based On Previews",
• "Beware the Histogram",
• "How to Trash a Good Shot in One Step...",
• "Why Bother Shooting RAW if Culling JPEGs"

But let's continue to evaluate shot 4630...

... Using a RAW Converter with Default Settings

If you don't use JPEGs for culling RAW, and don't yet use FastRawViewer for culling and selecting your RAW images :(, you probably opened this RAW shot directly in some raw converter. So did we. We decided to open this shot in ACR, using the default settings.

Shot 4630 opened in ACR with default settings: the image looks heavily overexposed (in R) and underexposed (in B). Is this indeed so?

It looks like this shot is very not OK - we still see some blob instead of red petals, which means that red channel is heavily clipped.

Checking for clipping areas for this shot might be tricky, because the clipping overlay has nearly the same color as this red flower. So here's a tip: switch the clipping warning ON, hold Alt and click-and-hold the "Exposure" slider. The whole image becomes black with red overlay on the clipped areas.

Clipped areas of shot 4630 according to ACR (with default settings)

Should we discard this shot as irreparably overexposed? Is it really that bad? What in fact do we see - is this a preview of RAW data recorded by a camera and clipping in the RAW data, or rather a preview and clipping in the "improved" image, resulting from applying white balance, Adobe baseline exposure (BLE), and a tonal curve to the RAW data? Let's find out...

...Using FastRawViewer

Finally, we open this shot in FastRawViewer and look at the unmodified RAW data. Is this even the same shot?

Shot 4630, opened in FastRawViewer. RAW Histogram and RAW exposure statistics panels in left top corner

As you can see, the red flower now looks quite normal, with all of the petals in place, properly detailed, and not melting together. Also, both the RAW histogram and the RAW exposure statistics, based on pure RAW data, confirm that there are no significant problems with this shot, neither in the highlights nor in the shadows (Raw Histogram and Raw Exposure Statistics panels are in the top left corner of the picture above, outlined by a red rectangle).

For viewing such shots, we recommend using the following settings in the FastRawViewer Preferences:

1. Do not check "apply Adobe hidden exposure correction" for image display;

FastRawViewer -> Preferences ->Image Display -> Exposure -> Uncheck box "Apply Adobe hidden exposure correction"

2. Set Contrast Curve to Gamma 1.8.

FastRawViewer -> Preferences ->Image Display -> Contrast Curve Type - Gamma 1.8

Now, if we find that the level of detail is still insufficient, we can quickly check the details in each channel separately, optionally with the Shadow Boost enabled, to get an idea as to what the level of details in each channels is, and what the level of noise is; in other words, whether we can use the Channel Blending technique suggested by Dan Margulis.

FastRawViewer. Shot 4630, channel R - we can see that there is plenty of detail in the red channel

FastRawViewer. Shot 4630, channel G; Shadow Boost set to 0.7

FastRawViewer. Shot 4630, channel B; Shadow Boost set to 0.5

Tuning a Raw Converter

If we want ACR to render the RAW such that it more accurately reflects what is actually there, we need to override some of the default "improvements" applied by ACR.

• Let's change some settings:
• Change color space from sRGB to ProPhoto RGB (number 1 on the picture below);
• Switch to Process Version 2 (2010 in older nomenclature) (number 2 on the picture below)
• Change the curve type to Linear (number 3 on the picture below),
• Set all values to zero (Blacks, Brightness, etc) (number 4 on the picture below);
• Override the baseline exposure correction: set the Exposure value to the negative value of BLE (you can find this value in "BLE" field of EXIF panel in FastRawViewer) - if the BLE is -0.5, set it to 0.5, and vice versa. In our case, it's 0.35, so we set it to -0.35.

And finally we can see the image that we have already seen in FastRawViewer, without, including according to the histogram, any of the false problems in the red channel that we saw with the default settings.

Shot 4630, ACR default settings are overridden

Let's look at the clipping areas now - nearly nothing is clipped:

Shot 4630, after ACR default settings is overridden real clipping is at minimum

Setting recovery to 25 easily deals with these tiny clipped spots on the flower stamens.

Shot 4630: clipped spots (left photo); recovery set to 25 (right photo)

So as you see, very often there is nothing wrong with your red flower shot, and you've simply been looking at it through a funhouse mirror.

To see what you actually got, you can use FastRawViewer or you can tune a raw converter to display RAW images without its unwanted (in this case) default "improvements."

• You can read more about tuning previews in raw converters in our articles:
• "Forcing a Raw Converter to Render Tones Accurately"
• "Deriving Hidden Baseline Exposure Compensation Applied by a Raw Converter"

However, we feel it necessary to ask: "Is it Worth it to Cull with a RAW Converter?"

P.S. Oh, and by the way, what would happened if, mislead by the in-camera histogram, we'd decreased the exposure while shooting this red flower; that is, we decided to go with conversion from shot 4625?

As was said above, shot 4625 has a nearly perfect _in-camera_ histogram, and has exposure 1 2/3 EV below the exposure of shot 4630. According to the JPEG histogram of this shot, the red channel is still a little bit clipped.

Shot 4625, embedded JPEG still has a little bit of clipped in the red channel

Here is what we have in the RAW for this shot - 18% of underexposed pixels in the B channel, 4.8% of underexposed pixels in the G channel and... 0.19% of underexposed pixels in the R channel (please have a look at the RAW Exposure Statistics).

Shot 4625, RAW is significantly underexposed

Here we can see the underexposed areas of this shot (highlighted in blue):

Shot 4625, RAW, underexposure overlay

And as they say, with much underexposure comes much sorrow; the more underexposure, the more noise...