The handling of highlights in negative conversions with ColorPerfect
Some users have asked about how the highlight compression system works in ColorPerfect. Also there frequently is the question why highlights would clip without it in the first place. We have further improved the highlight compression in ColorPerfect 1.04 and will try to answer these questions in the following article. To do so we'll look at a few basic characteristics of negative film first and at an example image thereafter to explain the compression system. This text is being written for photographers and assumes that you are familiar with photographic concepts like the combination of shutter speed and aperture resulting in a certain level of exposure.
How many exposure values can be captured on a single color negative?
In this first section we are going to see that negative film in general can capture more brightness information than can be represented in a positive image without resorting to special techniques. To give an initial impression on how large a tonal range can be represented in an average positive image let me quote from The Negative by Ansel Adams: "The range of zones which convey definite qualities of texture and the recognition of substance is the textural range, from Zones II to VIII." It is enough for us to know that in taking a photo one exposure value (EV) relates to one of the Zones Adams speaks of. The textural range of the positive image thus roughly spans a range of 7 stops. To also give an initial impression on how large a tonal range can be captured on negative film let me quote from the same book again: "The densities relating to exposures above Zone IX can be recorded, and in fact, considerable separation may exist on the negative for Zones X, XI, XII, and even higher." While the black and white films Adams speaks of may have even greater latitude than the average color negative film does we can safely assume that most color negative films also can capture a range of 10 or even more stops.
In the following paragraph we'll look at the characteristic curves of a recent color negative film to verify the above. It is of little importance but in case you wonder these curves belong to Kodak's Ektar 100. We'll try to keep things simple but should this become too technical for you to be comfortable with just skip ahead to the next paragraph explaining the actual highlight compression system in ColorPerfect.
Characteristic curves are diagrams of density versus LOG exposure. That means they show us how dense (dark) a given point on a negative will become when it is being exposed to a certain amount of light. It is not important to understand this conversion to density and its implications in detail. It is enough to know that the brightness and color information of a scene is stored on a negative in terms of density and that ColorPerfect can reverse this process to create a physically correct positive image of the scene again.
We want to verify how many stops or EV can be recorded on this color negative film while maintaining separation and detail. The LOG exposure scale in combination with the curves will roughly tell us. To interpret that data it's sufficient to know that an increment by 0.3 LOG Exposure is equivalent to an increment by one stop of exposure. We can put a grid of one stop exposure increments over the curves. Our leftmost line is placed before any noticeable increase in density occurs on the film. It thus represents pure black in any positive image. We'll assume the rightmost one to represent pure white even if there might be information beyond it. The curves don't simply stop where they do in the diagram but since the shoulder region of the curve is not plotted here we'll ignore it. The range between the two lines contains all tonal values a negative on this film can store in a differentiated way.
To better grasp what this means for a final positive image let's imagine our scene to be a simple gray scale. Starting from pure white on the right brightness will be halved with each stop. If we look at the left end of our gray scale we will find that there are several stop patches that appear to be black on the screen. A normal positive image will not show much discernible detail in these regions. If we want to discern detail in those parts of the image we will have to increase the general exposure level of our final positive. Now we can differentiate steps we couldn't before but in the process two stop patches to the right have become white. Not all negatives ever reach the maximum density level possible which means that they might contain significantly less information than we have visualized here. You may not have this much unprintable information but you will most often have some. The fact that negatives generally can store more information than can be printed as a positive image without using special methods answers the question why there might be clipping in the highlight region or undiscernible tones in the shadow regions of an image.
Using ColorPerfect's highlight compression system
Let's get started with ColorPerfect's highlight compression system by converting the linear scan of the negative to the right. First let's create an inversion that maintains all highlight information without applying any compression. The rightmost field in ColorPerfect's Highlight area is called Clip. It puts out the actual percentage of pixels that will clip in any color channel using the current settings. This read out always relates to the current crop of the preview image. If the lowest zoom factor is used it represents the entire image. To convert an image without any clipping we will have to add so much black that this read out becomes 0 or nearly so.
When doing that the image at hand looks too dark even though there still is some pure white in it in the specular highlights on the pearls in the flower bouquet. Look at the image and remember where the brightest regions on the wedding gown are. Those are the areas to compare between the following example adjustments. To come back to what we looked at theoretically above let's remove enough black from the image to roughly brighten the positive by 1 stop. Doing so will significantly raise the percentage of pixels that do clip without compression.
The highlight compression in ColorNeg always has emulated the compressive properties of analog photographic media which before were largely absent in the digital realm. ColorPerfect gives you more control in the process of this than ColorNeg ever did. There are two basic controls to use. The first is the count of stops to compress. We brightened the image by about 1 stop so if we compress that entire stop no clipping will occur at all. The second control lets you specify the range of tonal values in the final image that will be involved in this compression. If we choose the most narrow range of 250 for our image no clipping will occur but the level of detail preserved will be very limited due to that range. There will be little texture to the compressed tones on the gown. No compressed pixel will become darker than 250 in the color channel(s) that otherwise would have exceeded 255. Using a wide range for the compression like 200 will preserve more detail and thus yield more texture. The downside of that is that if a stop count is selected that will undo all clipping the image can appear to have less contrast after the compression is applied. In the example produced this way there don't even seem to be any specular highlights left. Most images benefit from a limited amount of pure white in them just as is the case for pure black. Compressing but 0.5 stops to a range of 200 works much better for this image in our opinion but that is a matter of preference. Without the compression almost 10% of all pixels would have been clipped. Our last settings bring this down to 0.6% so that the specular highlights are preserved. Testing the stop settings sequentially will tell you how many stops beyond 255 exist at your current settings. As a general rule of thumb you should find the lowest stop count that works well for you with any of the ranges and without causing too many pixels to clip. Some clipping often is acceptable. Do rely on your visual judgement - you are editing images, not histograms.
>In the example shown we removed a large amount of black to produce a meaningful example. In actual work there are numerous options to limit the amount of clipping that would occur without resorting to the highlight compression. We could have removed less black and could have used a Zone curve to brighten certain tonal regions more than others. We could have used ColorPerfect's unique tonal mode to change the luminance level of the skin tones or we could have used ColorPerfect's Gamma tool that also does preserve natural colors. We could have used advanced techniques involving selections to apply differing adjustments to different regions of the image. To produce the example images shown we did none of that. They are just basic inversions of a negative using a characterization built into ColorPerfect for the film type at hand.
Useful techniques for special images
If you require more control you can always work with multiple inversions. To do so copy your linear negative scan and paste it onto one or more other layers. Then convert the negative on the first layer using ColorPerfect to suit your needs in terms of overall image brightness, color balance, level of saturation etc. Do enable highlight compression in this conversion with settings that seem to work best. When you are done o.k. out of the plug-in select the next layer and start ColorPerfect again. On the start panel use the Initial>Previous button to restore all previously used adjustments in ColorPerfect. Now you can easily create a second inversion within seconds. E.g. by adding enough black so that all highlight detail is preserved. Afterwards you can blend the two or more layers together in Photoshop using layer transparency, the eraser and the Edit > Fade command. We advise you to stay away from both the Curves and the Levels dialogue in Photoshop for major adjustments because both are very likely going to impair your image's colors. Why and how they do is too much for the scope of this text and will be treated elsewhere. Do use ColorPerfect's TouchUp mode for any further changes to your image.