Optimizing Dynamic Range in Photoshop -- Part II

Article and Photography by Ron Bigelow

www.ronbigelow.com

Photoshop CS3 (Beta) Used in this Tutorial

Different Methods of Determining Dynamic Range

Often, different sources will list different dynamic ranges for the same camera or type of sensor. This is because there is no universally accepted method of determining dynamic range. So, different sources list different dynamic ranges because the sources are using different methods of determining dynamic range. There are three general approaches to determining dynamic range.

Calculated Approach: The first approach to determining dynamic range is to calculate it. This approach determines dynamic range by dividing the largest signal a pixel can produce (when the pixel has reached full capacity) to the noise that the pixel produces when it is receiving no light (primarily caused by dark current noise). While this approach has a nice mathematical definition, it has little relationship to what a photographer experiences when using a digital camera. In Part I of this series, dynamic range was defined as "the range from the darkest to the lightest tones that maintain accurate detail". As explained later in the article, the dark end of the dynamic range is partly determined by the SNR and partly by the human visual system's ability to distinguish detail in dark areas. This approach only accounts for the SNR portion of the dynamic range and ignores the human side of the issue. Consequently, this approach often yields dynamic range numbers much larger than what a photographer will be able to use in actual practice.

Measured Approach: The second approach to determining dynamic range is to measure it. In this approach, a target (often a graduated grayscale target) is used. The target has gray patches of different densities. Once photographed, the files are imported into a computer and the patches measured. The lower end of the dynamic range is determined by identifying the darkest tone that has not been clipped. The upper end of the dynamic range is determined by identifying the lightest tone that has not been clipped. This is a reasonable approach. However, it also does not consider the human eye's ability to distinguish detail.

Eyeball Approach: The third approach to determining dynamic range is to use the human eye to evaluate it. One way to do this is to photograph a target (again, this is often a graduated grayscale target) and view the target to determine which tones can be seen with the eye. While this is the least scientific approach, it yields dynamic range numbers that are closest to what a photographer experiences when using a camera. After all, photographers are really interested in the tones that they can see and that maintain detail, not what can be calculated or measured in a computer.

The Rest of this Series of Articles

Now that the basics of dynamic range have been covered, the rest of this series will focus on optimizing the dynamic range. The first thing that must be understood is that the useable dynamic range is determined by three items: the sensor, the system that is used to process the data from the sensor, and the human eye. For those that shoot raw, the system that processes the data from the sensor is the raw converter. Thus, the issue becomes "Which settings in the raw convert or will produce the largest dynamic range?" This is the focus of the rest of this series.

Method Used to Determine Dynamic Range

After all that has been said, you are probably wondering how the dynamic range was determined for this series. This series used both the measured and eyeball approaches. In both cases, a graduated grayscale target was photographed. For the measured approach, the raw file from the target was converted and measured in Photoshop. For the eyeball approach, the converted file was viewed on a monitor to determine which tones could be seen by a person viewing the monitor. Why two approaches? Two approaches were used because this series deals with two issues: 1) which settings in the raw convert or produce the largest dynamic range, and 2) what is the usable dynamic range that results from those settings?

The measured approach does a better job answering the first question. This approach produces consistent, quantifiable numbers. With this approach, it is not important to determine the exact dynamic range that is usable for a photographer. Rather, what are of interest are the relationships between the dynamic ranges that are produced by the different raw converter settings. Basically, we want to know which settings produce the largest dynamic range.

The eyeball approach does a better job answering the second question. The eyeball approach gives a better idea of the useable dynamic range that a photographer can expect in daily use. However, this approach is not as consistent as the measured approach. While viewing the same files, different people might come to different dynamic range numbers due to the differences in the visual systems from one person to another.

For these tests, a Transmission Stouffer Step Wedge target (see Figure 1) was photographed. This target is a piece of film with a series of gray patches (i.e., wedges). The wedges run from very dark tones on the left to very light tones on the right. Adjacent wedges vary in tonality by 1/3 of a stop. The entire target covers a range of thirteen stops. The target was sandwiched between two sheets of black poster board that had a hole just big enough for the target. The target was illuminated from behind by a very uniform light source. The area between the target and the camera was surrounded by black material to remove any reflections or extraneous light sources. The exposure was set for the middle gray tones.

Figure 1: Stouffer Transmission Step Wedge

The images were shot with a Canon 5D in raw mode. The raw files were converted in Adobe Camera Raw CS3 Beta.

For the measured approach, the brightness of the wedges was measured with the Eyedropper tool and the Info palette. The lower end of the dynamic range was set at a brightness reading of 2%. The upper end of the dynamic range was set at a brightness reading of 98%. In addition, if there appeared to be a problem with noise in the darker tones, the noise was checked by taking five measurements on each tone. The lower end was then defined as the darkest tone that had a brightness of 2% or greater and all five measurements were within 5 units of each other. Tones with greater than five units between the five measurements were considered to be too noisy.

For the eyeball approach, the wedges were viewed on the monitor. The ends of the dynamic range were determined by those wedges where tonal differences could be seen between adjacent wedges and noise did not appear to be a problem. It is important to understand that there can be a problem with the eyeball approach. Not all monitors can differentiate between shadow tones down to a brightness level of 2% or highlight tones up to a brightness level of 98%. If this is the case, the monitor will not show tones that actually exist in the file. This would result in an artificially low eyeball dynamic range. However, the monitor used for this test was shown to easily resolve tones to both the 2% and 98% brightness levels.

For each conversion, the upper and the lower ends of the target are shown. While larger differences in dynamic range may be seen in these images, subtle differences may be difficult to see as some detail was lost in preparing the images for the web. Thus, the original files in uncompressed TIFF format are available for download for closer evaluation.

JPEG

Before the work with the raw files began, the target was photographed with the camera in JPEG mode. Thus, the raw files could be compared to the JPEG image.

Figures 2 -- 3 show the lower and upper ends of the target for the JPEG file. Figure 4 shows the dynamic range data.

Figure 2: Lower End of Target for JPEG
Figure 3: Upper End of Target for JPEG
Figure 4: Dynamic Range Data for JPEG
Measured Approach
Lower End Wedge
Upper End Wedge
DR in Stops
33
8
8.33
Eyeball Approach
Lower End Wedge
Upper End Wedge
DR in Stops
28
7
7.00

The first thing that should be noticed is the significant difference in the dynamic range between the measured and eyeball approaches. The measured approach yields a rather impressive dynamic range of 8.33 stops while the eyeball approach yields a more conservative 7.00 stops. As can be seen in the table, both methods yield close to the same result at the upper end of the dynamic range. However, they differ at the lower end. Actually, this is not surprising. What this is demonstrating is that the computer can measure the differences between dark tones that the eye can not distinguish. The computer indicates that values 29 -- 33 have lightness values greater than or equal to 2%. However, to the eye, they all appear black.

So, which of these two dynamic range values is more reasonable for a photographer that wants to take pictures and produce prints? For this, the eyeball number is more realistic. A photographer could expect to produce tones in an image between wedges 28 and 8. Below wedge 28, it might be possible to use a steep curve to bring out the differences between the tones, but there would probably be noise problems and the image quality would not be good.

Full Size Image: JPEG

Raw Default Settings

The first raw conversion was done with the Camera Raw default settings.

Figures 5 -- 6 show the lower and upper ends of the target for the Camera Raw Default settings. Figure 7 shows the dynamic range data.

Figure 5: Lower End of Target for Raw Default Settings
Figure 6: Upper End of Target for Raw Default Settings
Figure 7: Dynamic Range for Raw Default
Measured Approach
Lower End Wedge
Upper End Wedge
DR in Stops
30
9
7.00
Eyeball Approach
Lower End Wedge
Upper End Wedge
DR in Stops
29
8
7.00

So, here we have a surprise. While the eyeball approach gives the same dynamic range value as the JPEG image, the measured approach gives a much smaller value for the converted image than the JPEG image.

Some people believe that shooting raw will give them a larger dynamic range than if they shoot JPEG. However, this case shows that this is not always true. While a raw file has a larger dynamic range than would be achieved if the scene was shot in JPEG, the raw conversion may produce a file with the same, or even less, dynamic range than the JPEG. It all depends on the settings that are used during the conversion. This shows the importance of identifying the best settings for a conversion rather than just accepting the default settings.

The next step is to identify the raw converter settings that will increase the dynamic range of the converted image. This conversion will serve as a baseline against which the other conversions can be compared.

Full Size Image: Default

Conversion Tools

Figure 8: Camera Raw Basic Dialogue Box

There are several controls in Camera Raw that can affect the dynamic range. Most of them are found in the Basic dialogue box (see Figure 8).

Figure 9: Camera Raw Tone Curve Dialogue box

The last control that will be considered is the Curve on the Tone Curve dialogue box (see Figure 9).

Tone Curve: The Tone Curve control determines which curve will be used to lighten and increase the contrast of the raw file during the conversion.

The Procedure

Each of these controls was adjusted separately. When one control was adjusted, all of the other controls remained in their default settings. The lower and upper ends of the target are shown for each conversion. The ends of the target for the default conversion will be repeated so that the reader can compare the results of the change in the setting. The results are shown in the associated table.

While this approach will be used as a starting point, it has a major flaw. It only tells what happens when one control is changed while all of the other controls are kept in their default settings. However, the results might be different if the other controls were not in their default settings. This problem will be resolved in Part IV of this series. Meanwhile, the data covered in Part II and Part III is very important because it helps us begin to develop an understanding of the impact of each setting. In addition, this data is needed to set up the more advanced analysis that will be performed in Part IV.

Exposure

The first control to be adjusted was the Exposure control. This controls the exposure of the converted image and attempts to extract detail from the highlights in areas were one or two of the channels are clipped (when the Exposure setting is set to a negative value.) In addition to the default setting, conversions were done at -1.50 and +1.50.

Figures 10 -- 15 show the lower and upper ends of the target at the three settings. Figure 16 shows the summary data.

Figure 10: Lower End of Target for Default Exposure Settings
Figure 11: Upper End of Target for Default Exposure Settings
Figure 12: Lower End of Target for Exposure Setting of -1.50
Figure 13: Upper End of Target for Exposure Setting of -1.50
Figure 14: Lower End of Target for Exposure Setting of +1.50
Figure 15: Upper End of Target for Exposure Setting of +1.50 
Figure 16: Dynamic Range for Exposure Control
Measured Approach
 
Lower End Wedge
Upper End Wedge
DR in Stops
Default 
30
9
7.00
-1.50 
29
5
8.00
+1.50 
28
13
5.00
Eyeball Approach
 
Lower End Wedge
Upper End Wedge
DR in Stops
Default 
29
8
7.00
-1.50 
28
5
7.66
+1.50 
28
13
5.00
For both the measured and eyeball approaches, it is obvious that decreasing the exposure increased the dynamic range. Basically, very little was lost in the dark tones, but tones were gained in the highlights. Conversely, increasing the exposure significantly decreased the dynamic range due to two factors. First, there was a very significant loss of detail in the highlights. Second, there was a noise problem in the shadows. For the measured approach, a brightness level of 2% was reached at wedge 31, but wedges 29 -- 31 failed the noise test. So, rather than increasing shadow detail, increasing the exposure actually lost detail in the shadows due to noise. It is possible that increasing the exposure by a smaller amount might have done a better job with the shadows.

Full Size Image: Default

Full Size Image: Exposure -1.5

Full Size Image: Exposure +1.5

Recovery

The next control to be adjusted was the Recovery control. The Recovery control is used to recover highlight detail. In addition to the default setting, conversions were done at settings of 33 and 66.

Figures 17 -- 22 show the lower and upper ends of the target at the three settings. Figure 23 shows the summary data.

Figure 17: Lower End of Target for Default Recovery Settings
Figure 18: Upper End of Target for Default Recovery Settings
Figure 19: Lower End of Target for Recovery Setting of 33
Figure 20: Upper End of Target for Recovery Setting of 33
Figure 21: Lower End of Target for Recovery Setting of 66
Figure 22: Upper End of Target for Recovery Setting of 66 
Figure 23: Dynamic Range for Recovery Control
Measured Approach
 
Lower End Wedge
Upper End Wedge
DR in Stops
Default 
30
9
7.00
33
30
6
8.00
66
30
6
8.00
Eyeball Approach
 
Lower End Wedge
Upper End Wedge
DR in Stops
Default 
29
8
7.00
33
29
5
8.00
66
29
5
8.00
The interesting thing with this control is that, for both the measured and eyeball approaches, increasing the setting from 0 (the default) to 33 resulted in a dynamic range increase of one stop. However, increasing the setting to 66 caused no additional gain.

Full Size Image: Default

Full Size Image: Recovery 33

Full Size Image: Exposure 66

Articles

Dynamic Range -- Part I     Dynamic Range -- Part III