The Three Most Obvious Reasons to Look at RAW and Not Cull Based On Previews

A typical argument you might find on a photography forum:

"...Really, why do you even want to look at RAW files? The whole point of RAW is to be processed according to your taste into a JPEG. I never look at RAW files; I never need to. They are loaded into LR, processed, and I look at the processed images.”

There is no question as to which is more informative - the JPEG preview (left) or the RAW (right)

Let’s gloss over the part of whether it is possible to see RAW at all. If you want details on this subject you can read our article Dispelling a Myth: Viewing RAW is Impossible. In that article, we just explain that viewing RAW is no less possible than viewing JPEG or TIFF. Of course, none of them are directly viewable, except as hex dumps, and that doesn’t really count as viewing, now does it? So, the option of viewing or not viewing RAW really depends on the software that you use. Most pieces of software can’t or don’t display RAW – mostly because they can’t do it quickly; some, such as FastRawViewer and RawDigger can and do.

Back to the initial question – why do you want, or, as a matter of fact, not just want but need to look at RAW and not some previews rendered in some arcane manner when you’re choosing RAW files for conversion or presentation? Why, in fact, shouldn’t you trust neither embedded nor rendered JPEGs/previews, nor the preview / histogram on the back of your camera, for that matter?

We are going to both answer these questions and illustrate the massive disconnect between a preview and real RAW data.

Unfortunately, photographers are throwing away perfectly good captures while keeping captures that are sub-optimal, because they are not being provided with the facts about the RAW. We are going to demonstrate why they need to examine actual RAW data before making any decisions about which captures to keep or consider editing.

Here come the answers.

Reason №1: previews do not allow one to adequately assess the exposure of the RAW shot

When we say “previews”, we mean those renders “normally” used for evaluating and selecting RAW shots: embedded and separate out-of-camera JPEGs, previews on the back of a camera, as well as previews rendered by a raw converter. All the previews, no matter how they are created, are “improved” in un-clear ways, having their brightness adjusted, white balance and tone curve applied, shadows plugged, highlights “recovered”, and are in no ways representative of the RAW data.

Consequently, previews are prone to the following:

1. Two types of false clipping – that is, prominent clipping in JPEG when there is no clipping in RAW data

1. False Highlight clipping in JPEG for optimally exposed RAW
2. Because of the brightness adjustments, white balance, and tone curves applied while converting RAW, the JPEG data might be so much “hotter” than the original RAW, that the JPEG, automatically converted from RAW with optimal exposure for RAW, will falsely indicate severely clipped highlights.

Figure 1. Shot 1866, a beach in Greece at noon – highlight clipping in out-of-camera JPEG and optimal expose in RAW:

Figure 2. Marigold: highlight clipping in out-of-camera JPEG and optimal exposure in RAW

2. False clipping in one or two channels (red and blue) in the JPEG because of white balance being applied during RAW conversion
3. RAW data recorded by a camera is not affected by the application of white balance because white balance is not applied to the RAW data but recorded separately as metadata. On the other hand, applying white balance is one of the necessary steps that will affect the final result of regular RAW conversion, whatever it may be (both in the camera and in a RAW converter).
4. Technically, white balance application is the multiplication of RAW data values for the red and blue channels by corresponding white balance coefficients; the values of these coefficients nearly always exceed 1 and sometimes reach 4+. This multiplication, increasing the values in the red and blue channels during the process of uncontrolled RAW conversion, shifts the histograms of those channels to the right and might cause the channels to reach clipping point. Meanwhile, in RAW, both the red and blue channels might be perfectly OK.
5. White balance is nothing more than per-channel exposure correction, and it causes clipping the same way general excessive linear exposure correction does.

Figure 3. Shot 2171, Pink Azaleas: clipping in JPEG due to white balance application;

white balance (R,G,B) coefficients are (1.66, 1, 1,79) – follow the red arrows at the bottom bar:

Figure 4. Red Lilies – clipping in JPEG due to white balance application; white balance coefficients are (1.34, 1, 1.73):

Figure 5. Red Tulips: clipping in JPEG due to white balance application; white balance coefficients are (1.84, 1, 1.66):

Figure 6. Orange Irises: clipping in JPEG due to white balance application; white balance coefficients are (2.52, 1, 1.54):

So, we’ve demonstrated how one, being misled by an embedded or rendered JPEG/preview, may throw away optimally exposed RAW shots. Now, let’s see how such a preview may convince one to choose to keep suboptimal RAW shots:

2. False “hottest possible” exposure – seriously underexposed RAW might be rendered as a “perfectly exposed” JPEG.

Because of the brightness adjustments, white balance, and tone curve applied while converting RAW, a JPEG might become significantly “hotter” than the original RAW.

Figure 7. Shot 1861, same beach in Greece as on fig.1 – hottest possible exposure for JPEG, while RAW remains significantly underexposed:

Using RAW-based tools such as a RAW Histogram, RAW Exposure statistics, and RAW-based Overexposure Indication together with the Exposure Correction tool, we can see that this shot was underexposed by 1 2/3 EV (some small overexposed areas on the pebbles start to appear only after applying +1.87 EV Exposure Correction to this shot)

Figure 8. Shot 2175, Azaleas – hottest possible exposure for JPEG, while RAW remains significantly underexposed:

So, to assess the exposure properly, one needs a histogram and exposure statistics based directly on RAW data.

FastRawViewer allows you to evaluate the exposure of the shot by examining on-the-fly RAW conversion done without any brightness adjustments applied behind the scenes, and by looking at the real RAW histogram, RAW Exposure statistics, and Over- and Underexposure overlays based directly on the RAW data.

You can also determine the amount of underexposure using the Exposure Correction tool. To do so, start increasing the Exposure Correction value (see Figure 9, marked in red at the bottom bar of the main window of FastRawViewer) until the OE+Corr column in Exposure Statistics panel indicates a percentage of overexposed pixels in the strongest channel that is slightly higher than zero, but less than 0.1 - 0.3% (outlined in red in the Exposure Stat panels). The Exposure Statistics (the OE+Corr column) will tell you how many pixels and/or the percentage of pixels that would have been blown out as a result of hotter exposure, while an overexposure overlay will highlight the would-be blown-out areas.

Figure 9. RAW view: “as shot” and after applying +1 EV Exposure Correction in FastRawViewer

By the way, on a real shot that was exposed 1 stop hotter, the number of underexposed pixels is much decreased.

Some would claim that by tuning the camera settings one might achieve “correct” histogram. To do so, one needs to load into a camera specialized custom white balance setting, UniWB, and a curve that pulls the midpoint approximately 1 stop down. As a result, previews become dark and in funny green color. You can read more on the subject in our article “Beware the Histogram”

Reason №2 looking at a preview, it's hard to determine which details are retained in highlights

Very often, previews demonstrate flat featureless areas and blobs in highlights, and looking at those previews it is very difficult to decide if it is possible to do something with those lifeless areas during the RAW conversion stage. Direct use of raw converter to determine that takes a while: one needs to play with different sliders and settings to render a preview that allows evaluation of the highlights.

FastRawViewer offers a Highlight Inspection tool, accessible quickly through the single keystroke (Shift-H), with the purpose of immediately bringing back those seemingly void highlights without applying any algorithmic (and thus somewhat misleading) highlight recovery. That is, Highlight Inspection in FastRawViewer displays the true content of highlights in RAW.

Figure 10. Cherry Blossom and Overcast Sky – are there any details?

Figure 11. Daffodils and Overcast Sky:

Figure 12. Cherry Blossom Alley:

Figure 13. Meteora (Kalabaka), Greece

Reason №3: a preview often doesn't allow one to evaluate shadows for details and noise

Very often, previews display completely plugged featureless shadows, while RAW conversely has enough details and not too much noise in the shadows, allowing one to get very good results during RAW conversion. This is a serious issue with shots of scenes with high dynamic ranges with extremely important highlights (such as wedding shots with the bride in a dazzling white gown and the groom in a black tuxedo) or nighttime/late evening shots when it is simply impossible to increase the exposure (because of the danger of severe clipping of the highlights, like what happens with sunsets, night street photography, concerts, etc.) but the scene is worth having its picture taken.

Once again, one can play with the sliders and settings directly in a RAW converter, rendering the previews with opened shadows.

The Shadow Boost tool in FastRawViewer allows one to open the shadows with one keystroke (Shift-S).

Figure 14. Passing Small Village Wedding, someplace in Europe:

Figure 15. Central Park at sunset, NYC

Figure 16. At night near the Louvre:

Figure 17. Evening in Greece:

Figure 18. Yellowstone, Mammoth

FastRawViewer additionally allows one to inspect the noise in the shadows using the Focus Peaking tool in Details mode (press P two times – pressing P one time turns on Focus Peaking’s Edge mode, which allows one to see the sharp contours and edges of the image; the second press turns Details mode on).

Figure 19. Bridge in Zagori, Epirus, Greece:

Figure 20. Inside of the National Cathedral, Washington DC:

Figure 21. Walls of the National Cathedral:

Additionally, you can check the noise level per channel – R, G, or B. To examine the image, below we used Per-Channel View (Blue channel), and turned on Shadow Boost and Focus Peaking in the Details mode (Cmd-5, Shift-S, press P).

Figure 22. Just a rose

We haven’t even mentioned here that if one were using UniWB, it would be quite difficult to evaluate the shot based on its preview.

Figure 23. The shot made with UniWB

Also, if one has accidentally chosen a wrong white balance setting, or decided in favor of Auto white balance, and the camera miscalculated it, the out-of-camera JPEG (or a default render made by a raw converter, which picked the camera’s settings) will look rather “strange” and/or may demonstrate clipping due to white balance application.

The picture below presents two shots made with the same High Color Rendition (Day White) Fluorescent light (HiCR FL-N), but for the first shot the camera was manually set to Tungsten WB; for the other shot, the setting was Shade WB. Let’s look at the picture below – as you can see, both RAW views and both RAW histograms are absolutely identical, indicating optimal exposure for RAW. At the same time, the JPEG previews look very different, and the JPEG histograms for both JPEGs indicate severe clipping but in different channels – the blue channel for the first JPEG, and the red channel for the second JPEG. That is, we are faced here with the classic case of clipping caused by white balance application during in-camera raw conversion.

Figure 24. Two shots of ColorChecker SG under the same light but with different in-camera white balance settings:

We hope the above examples constitute a compelling argument supporting the proposition that one needs to look at RAW and examine the actual RAW data before making any decisions regarding which captures to keep or consider editing.

And, by the way, all of FastRawViewer’s tools that we used here can be accessed through simple keystrokes, and all of them work extremely quickly – much faster than the iterative rendering of previews through a raw converter.

You can try all of this for yourself for free – just download a fully functional 30-days trial of FastRawViewer (it’s exactly the same download as for those who buy FastRawViewer); also, we reply to everyone without regard as to whether you’ve already bought the program, or have simply decided to give it a try, and we welcome any feedback, requests, and suggestions provided to us.