Breakthrough Precise Frequency and Amplitude tracking.

A Better Method of Sharpening Pictures in Photoshop

This technique will be completely different than any picture sharpening technique you have seen before. Properly used, it will produce a picture that appears distinctly sharper when viewed from a normal viewing distance, but which will show no obvious sharpening artifacts, even viewed close up. In fact, if you do a close-up side-by-side comparison with small parts of the original the differences – even the added sharpness – will not be obvious. I currently use this technique on the majority of photos from which I intend to make large prints, and it is especially useful when trying to stretch the resolution of a picture to make a convincing large print. I have used it several times to make convincing prints covering most of 13" x 19" paper from originals that are 1000 to 1600 pixels in the major dimension. However, that should not be expected to work satisfactorily for every picture.

First, prepare the picture. Do any image content corrections that are planned, fixing scratches, etc. Do any gross corrections in levels, color, brightness, etc. that are necessary to bring the picture to the approximate overall final state for printing. I do not use layers to do these initial gross corrections, but if you do, flatten the image. Resize the picture to sufficient pixels to make a good print. My own nominal standard for this is 300 pixels per inch in the final print. If you normally do some "regular" sharpening, this is the time to do it. In fact, any regular sharpening that is not overdone will be enhanced by the method. I sometimes do this, but I recommend first trying the method without regular sharpening.

Now use the technique. First, make a duplicate layer of the Background layer. I usually call the duplicate layer "Sharp." With the duplicate layer selected, set up an extreme unsharp mask. Settings which might be considered typical for this step are: Amount 150%, Radius 250 pixels, Threshold 5 levels. Those are not typographical errors. They really are the appropriate settings for an image with a nominal 5000 pixels on the major side. For different pixel counts, try a Radius of about 0.05 times the maximum pixel dimension.

Unlike the normal usage of the unsharp mask, you will find this takes a very long time. You may have to wait many seconds or even minutes for a preview and even if you skip on waiting for that, it will take a long time to perform the filter function – again perhaps minutes. Do not be alarmed by the result. What you want at this stage is an image that appears very sharp although parts may be lacking in detail and will have apparent high contrast and patches of exaggerated color. About the only significant problem that I have seen at this point is wide, very light, halos around large dark areas of the picture, and I have only seen that once or twice. The proper way to solve that is to select out the dark area and separately sharpen the two parts of the picture, combining the results. But if this happens, select another picture for a first test of the method.

The next step is to change the opacity of the Sharp (or whatever you called the duplicate) layer. Select the Layers tab in the Layers-Channels-Paths box and select the Sharp layer. Make sure "Normal" is selected on the pulldown at top left of the box and in the top right of the box use the Opacity pulldown to select an opacity of 20%. That setting should be approximately correct, but you will need to experiment with setting it at values between about 10% and 40%, switching the Sharp layer on and off to compare with the original. I usually leave it at a point where the overall sharpening effect is obvious during on/off comparisons but close examination does not reveal any annoying artifacts. Usually this is 15% to 20%, but it varies a lot for different pictures. It is likely that at first you may select opacities that are a little too high for best print results. Once you are satisfied, you can merge the Background and Sharp layers, although I often leave them separate. Then continue on to make any further layers adjustments, masking, etc., to bring the picture to its final appearance. The sharpening process will increase apparent image contrast somewhat. Often this is beneficial, but if it is not, use standard adjustment techniques to remove it.

While it is necessary to experiment with the opacity setting to get the best result, you may also find it interesting to experiment with the unsharp mask settings. You may find that somewhat different settings work better with your setup. If you are interested in why this technique works, send me an e-mail. The technique has more history behind it than you might think. As this method of implementation is original with us, you are welcome to use it in an article, a course, or a book but please credit C F Systems and www.c-f-systems.com

Return to Tips List


Color Balancing Difficult Cases

When a color photograph is not close enough to proper color balance to tweak it in, it sometimes is very hard to home in on a good balance. This may be a routine problem if you work with old prints or negatives. In our experience the following sequence will work satisfactorily on most such pictures after you have used it a few times to get the hang of it. It works more easily on some pictures than on others, so try several different pictures if you have trouble getting it to work at first. As you try this for the first time, it may be hard to believe it works. Be patient until the color corrected image suddenly appears in step 7. Once you are used to it, the whole sequence will take less than a minute.

All of this is done using the Photoshop Levels control and RGB color. Since we assume the picture is well off proper balance and will require major adjustment, scan in a 16 Bits/Channel original and work the following in that mode if it is at all possible. 8 Bits/Channel will also work, but banding and other artifacts are more likely to appear in the final result.

With some pictures, especially from faded prints or slides, you may find that steps 1 through 3 have done the job and that only a little further tweaking is necessary. Other pictures may gain very little from these steps, but will respond to steps 3 through 8, and some pictures may require all the steps to become presentable.

1. First, select the Red Channel. If the levels graph tapers off to nothing at either or both ends, adjust as shown:

Move the left pointer to the right, to where the data start to show and then move the right pointer to the left to where the data start to show. Often on the right side there is a tail of data barely above zero (as shown above) and when that happens it is usually best to put the right pointer midway along the tail. When a tail is present on the left side it is often best to move the left pointer completely past the tail, to where the data begin to rise.

2. Do the same procedure with the Green channel and the Blue channel. Do nothing yet to the RGB channel.

3. Now select the RGB channel and if necessary adjust the center pointer so that the picture is just slightly darker than you would like.

4. Again, select the Red channel. This time move the center pointer far enough to the left so that the picture takes on a decidedly red tone:

Not just a red tinge, but way too red; red enough that a complete novice looking at it would remark on its being too red. The value of 1.99 shown above is only a rough guideline but should give the general idea.

5. Now select the Green channel. If you move the center pointer way to the left the picture will take on a sickly green cast. Move it way to the right and it will take on an ugly magenta cast. Slide it back and forth between the extremes and you will see that somewhere in the middle there is quite a sharp point of transition between too green and too magenta. Find that point and then move just slightly to the right of it so that middle tone neutrals or tans have a slight but definite magenta cast.

6. Select the Blue channel. If you move the center pointer way to the left the picture will take on a heavy blue cast. Move it way to the right and it will take on a heavy yellow cast. Slide it back and forth between the extremes and you will see that somewhere in the middle there is a point of transition between too blue and too yellow, but it will not be as sharp as the transition between green and magenta seen in step 5. Find the point of transition and then move just slightly to the right of it so that middle tone neutrals have a slight yellow cast.

7. Now select the Red channel again. Move the center pointer back toward the right. In most cases as the red decreases there will be a point where the color balance is acceptable or very nearly so. Because you start with a picture that is way too red there is a tendency to overdo the shift toward cyan at this point and you may wish to slide back a little toward red to compensate, just enough to make the picture "warmer." Even when the color balance is still not quite right it will usually be close enough that normal tweaking will work.

8. Select the RGB channel and adjust the overall lightness/darkness as you would like.

In general, there should be an improvement in the picture in Steps 1 and 2. If the picture actually appears worse after Step 2, try again with the original picture, starting at Step 3. If you have run the entire sequence and the final picture is too contrasty, try again starting at Step 1, but in Steps 1 and 2 move the pointers only half as far in from the ends.

As this method is original with us, you are welcome to use it in an article, a course, or a book but please credit C F Systems and www.c-f-systems.com.

Return to Tips List

Accuracy and Precision in Digital Photography

A technical discussion of this topic could easily fill a book, but everyone seriously using digital photography needs to know the basics. Our views may differ from others you have seen because we have extensive experience with mathematics and the practical adaptation of mathematics to the constraints posed by computer arithmetic.

You can find "precision" as a dictionary definition of "accuracy," but the words often have distinctly different meanings in technical work. They both refer to how exactly the real world matches some representation of it, but accuracy is used in a global sense and precision in a local sense. Photographic images are rarely accurate representations of the real world scene they correspond to. The darker regions of the photo have to be lighter than the real scene and the lighter areas of the photo have to be darker than the real scene because they are limited by the blackness of the ink (or image silver) and the whiteness of the paper. The various mid-tones often appear darker or lighter than a direct measurement of the real scene would call for even if the overall contrast range of the real scene has been compressed to the contrast range of the photographic print. As for color, the spectrum of light reflected by a color in a photograph normally is completely different than the spectrum of light reflected from the real object it represents even though the colors may appear very similar to the eye.

Precision is a measure of how well an image behaves in a local sense (a differential sense if you are mathematically minded). The image may differ from the real scene, but is the difference consistent locally? For example, do areas with nearly uniform color have gradients that appear as smooth as in the original scene? Where there are two adjacent objects, one slightly darker than the other, is the darker one also the darker one in the real scene? In the brightest areas of a scene, silver-based films start to get progressively less sensitive to light, so rather than being abruptly all white, the brightest areas grade into whiteness. This is the "s-curve" sensitivity of silver-based media and it results in very desirable increased highlight detail from this better precision – but less accuracy. Our white paper (CFS-211b) gives a technical explanation of how the decision to be more accurate at the expense of precision is a major problem with current digital cameras.

Even with a very inaccurate representation of colors, the way color relationships behave in the print may visually match the way they behave in viewing the scene, giving a convincing representation when viewed apart from the real scene. Side-by-side comparison of corresponding pieces of such a print and the visual scene may be remarkably different not only in the spectrum of light they reflect but they often will be markedly different visually. Books on color theory all give extreme examples of this. [We suggest: Roy S. Berns: "Billmeyer and Saltzman's Principles of Color Photography," Third Edition, Wiley, 2000] You can see for yourself by clipping out pieces of a print that you believe is accurate and comparing them side-by-side with the original scene.

It is important to understand the difference between accuracy and precision because in photography, precision is often of critical importance while accuracy is rarely important and actually can be detrimental. [Note: There are photographic applications where accuracy is more important than precision; photometry in astronomy and certain chemical analysis methods, for example, and advertising, where it can be important to exactly match logo colors. People practicing in such areas will already be aware of the distinction.]

The basic problems of precision in digital photography are more easy to understand by looking at B&W first. In digital B&W photography each pixel is represented by a number measuring the darkness of the gray in that pixel of the picture. In the standard mode Photoshop calls 8 Bits/Channel, the number runs from 0, meaning fully black, to 255, meaning fully white. These numbers must be integers, like 134, 135, 136. Intermediate values like 135.2 or 134-1/3 are not possible, so for the image to have visual precision it is important that the difference between the grays represented by any two successive numbers be smaller than the difference that can be detected visually. That can be easily tested: a strip which grades from black = 0 at one end, through 1, 2, 3 … up through 253, 254, 255 = white at the other end will grade gradually from white to black and will show no "steps" at any point in between. Are 254 shades of gray between black and white sufficient to do that?

The short answer is "yes." It is generally accepted that human eyes can distinguish about 100 levels of gray, as measured by the "L" scale that is used in the Lab and Luv international color systems. Clearly, 254 levels can be more than enough, but there is a problem. The human eye is more able to distinguish between subtly different dark grays than subtly different light grays, so although the L scale has 100 intervals, in terms of intensity the intervals are more closely spaced together at the dark end and more widely spaced at the light end. So, although it seems logical to divide the intensity range into 255 equal steps starting with black = 0 and ending with 255 = full white intensity, that scale would show stair-steps in the gradient at the dark end. To work properly, more of the 254 intermediate grays must be placed below middle gray than above middle gray and the logical-sounding scale with even intensity steps is not generally used. A legacy of television history called "gamma" has led to a standard in which the 254 gray levels have many more levels of dark gray than light gray. The gamma response does not exactly match the sensitivity of the human eye, but nearly enough so that the 256 gray levels provide good coverage of the 100 gray levels that the human eye can see.

There has been controversy surrounding gamma. It was originally developed for early television as an adjustment to a television signal so that an unmodified CRT used as a television picture tube would produce a reasonable gray scale and thus a good B&W picture. CRTs are still used as television and computer displays and the gamma correction has remained in use. It is a very remarkable coincidence that the gamma correction for a CRT nearly matches the behavior of the human eye. In fact, the gamma of 2.2 used in PCs is a closer match than the gamma of 1.8 which Apple uses. Apple's gamma of 1.8 came about from another legacy compromise required for a now-extinct printer. If you search the web for gamma you will find a lot from and about Timo Autiokari, who has developed an entire approach to digital color photography based on the use of the "natural" gamma of 1.0. This is the same as the equal intensity step scale mentioned above which we say here will not represent darker grays adequately if applied as 255 steps. There has been a lot of invective exchanged between Autiokari and others. Although Autiokari is clearly wrong about some key details, he just as clearly has a good technical handle on some other difficult areas of imaging and there are people having success using the system he has developed based on the an incorrect understanding of gamma. How that can be is at least partly explained below.

So, 8 Bits/Channel can represent a B&W photograph. The problem is when you work with an 8 Bits/Channel image, almost anything you do – including just printing it – will make it behave more like 7 Bits/Channel. Anyone working with the "Levels" control in Photoshop has seen this. Alter the levels and come back to the image and suddenly you will see a "comb" appearance in the levels graph. This means that many of the 256 levels no longer appear anywhere in the image. Additional or more extreme manipulations usually spread apart the "teeth" of the comb, meaning that there are sequential pixel values that appear nowhere in the picture. If half of the possible pixel values are missing, for example, the remaining 128 are barely more than the 100 required. Where gaps having more than one missing pixel value in a row start appearing, smooth gradients are impossible and "stair-steps" in evenly toned areas result.

So, working with photographs in 8 Bits/Channel is possible but very marginal. Anyone with much experience working in Photoshop will have seen the comb effects and the stair-step gradients that result. Yet people successfully work with 8 Bits/Channel images in Photoshop every day. How can that be?

The main reason that 8 Bits/Channel can be made to work is that almost all such work is done in color rather than B&W. Anyone who has worked with B&W in 8 Bits/Channel mode in Photoshop will realize that the problems described above are more difficult to deal with than when working with color. Color work typically uses three "channels" and thus three 8-bit values to represent each pixel. These three channels are each subject to the same problems as the single B&W channel, but the numerical problems, such as the comb effect, usually do not affect all three channels in the same way and in individual pixels the three channels tend to compensate for one another. What would be experienced as a visible intensity change when working in B&W may be an imperceptible color change, working in color. That is, although they may be present in each individual channel, the stair steps are less obvious or invisible when the channels are combined in a color image. Of course, in an area where one primary color greatly predominates, the stair-steps are more likely.

So the stair-step and blocky region effects are less likely to be a problem in color images even though they still exist.

Dithering

You should also be aware that without saying much about it, Photoshop also makes rather extensive use of another technique to hide the loss of precision that leads to the stair-step and blocky region effects. Photoshop calls the technique "dither," although it seems a rather expanded definition for that word. Photoshop applies this particular type of dither by randomizing the least significant bit of pixels as part of a calculation. That is equivalent to calculating a pixel value, say 135, and tossing a coin to decide whether an additional 1 will be added. Thus some pixels that calculate to be 135 will end up as 136. This does a good job of insuring that all pixel values are represented, so no comb or stair-step effect, but it does so by intentionally adding noise to the image. There will be no stair-step, but there also will not be the nice, smooth gradient that originally existed.

The most egregious misuse of this method is that it is used when converting 16 Bit/Channel images to 8 Bit/Channel, so after carefully working in 16-Bit to preserve precision, when you convert Photoshop does its best to nullify this by effectively giving you a 7-Bit image rather than 8-Bit. Prior to Photoshop 6, it was actually necessary to manually patch the Photoshop program code to stop it from doing this. In versions 6 and 7 you can turn off dithering. Edit>>Color Settings will show the following dialog. Note that "Advanced Mode" is checked to make the whole dialog box show. Under "Conversion Options" "Use Dither" should be unchecked to prevent the action described above. Photoshop's description of this is interesting: "Use Dither: Controls whether to dither colors when converting 8-bit-per-channel images between color spaces. Dithering greatly reduces banding artifacts, but may increase file size." Regardless of what it says, dithering shows up in the 16-bit to 8-bit conversions and in other places as well. The "increased file size" is what you would expect when you add noise to an image, making it more difficult to compress. It is also important to realize that if dithering is on, Photoshop will add dithering noise to any picture as it goes to the printer.


You will find people – some widely regarded as authoritative – who will tell you that using 16-Bits/Channel does not make a difference in the final result. These people – authorities or not – simply do not understand what they are talking about. They may be describing what they see, not realizing that a few operations having "dither" turned on may completely throw away what they have gained from the 16-Bit operations. The differences may be completely lost in a haze of noise that degrades the whole picture. Whatever the cause, there are many cases which call for operating in 16-Bits/Channel if at all possible, where it definitely will make a positive difference.

So what about Timo Autiokari's method that uses a gamma of 1.0? We stated earlier that 8 Bits/Channel and a grayscale based on uniform intensity steps – which is a gamma of 1.0 – will show stair-steps. Again, Autiokari's method benefits from the fact that it is intended for color images. While the eye distinguishes differences in intensity better among darker grays, the eye distinguishes color casts more easily in the lighter grays. With Autiokari's method, the three channels still tend to compensate for one another in the dark region where his gamma = 1.0 gives a problem, and the system has more latitude for handling color casts in the lighter regions. Couple this with Autiokari's general recommendation to do what you can in 16-Bit/Channel mode and there may well be no more problems with his system than with the more widely accepted methods. Certainly the problems should be different. Autiokari is well aware of the "dither" problem in Photoshop and in fact it was from his web site that we learned that Photoshop "dithered" – effectively added noise during – the conversion of 16-Bit/Channel images to 8-Bit/Channel images, a fact so incredible we might never have thought to look for it.

If you are trying to produce high quality prints or images, precision counts. Use 16-Bit/Channel whenever possible, especially for early edits that represent significant changes in the image. Convert to 8-Bit/Channel only once, after all major edits have been made, and do not use dither. [Hint: For easier and more comprehensive editing in 16-Bit/Channel mode, temporarily convert the image to 8-Bits/Channel and use full Photoshop features to make any selection masks you will need for editing control. Save the selection masks in a separate file. Return to your 16-Bit/Channel image and you will be able to load the selection masks from the separate file.]

Return to Tips List

Brightness, Contrast and Loss of Detail

Before delving into the subtleties of the Photoshop Brightness/Contrast control, another interesting brightness/contrast issue is worth visiting. It is not unusual for an experienced photographer to feel a bit uneasy in adjusting the picture quality of a television set or computer monitor. It gets done, but it is more difficult than it seems it should be, particularly for brightness and contrast. The reason for this is that from the very beginning television controls have been essentially opposite to what a photographer (and probably most other people) would expect. The brightness TV control adjusts mainly picture contrast and the contrast TV control adjusts mainly brightness. This is the reason that monitor adjustment procedures typically suggest that you set contrast to 100%, to get maximum brightness, and then set brightness to expand or contract the image contrast until the blacks just become black. TV displays are a lot easier to adjust once you know this.

The Brightness/Contrast control in Photoshop does operate as a photographer would expect. The effect is so similar to what the photographer expects from working with silver-based materials that he may be trapped into accepting the subtle way in which the control steals highlight and shadow detail that silver-based methods would not. Here we will use a B&W example, but the same principles applies to color.

The following is a scan of an old photo and its RGB "histogram" as displayed by the Levels control. The photo is of my father and his brother, circa 1910, and was a 2" by 3" print of moderate to low quality. This example was chosen not because the stealing of highlight and shadow detail will be extreme or even especially important to the picture, but to illustrate that it can show up noticeably even in lower quality images like this one. Download a zip file (about 1 MB) with a 16 Bits/Channel Tiff file of the picture and the loadable Levels and Curves adjustments files if you want to trace the following steps for yourself.

There are no pure whites or pure blacks in the picture and the histogram shows this as an odd shaped lump in the middle grays with nothing at the light and dark ends, meaning there are no pixels at all that are very light or very dark gray. For reasons we hope will become clear later, we typically target 16-bit scans to appear roughly like this, although we normally try to make the histogram somewhat wider.

Now, we use the Photoshop Brightness/Contrast control to adjust the picture – here we will use the Image>Adjustments>Auto Contrast control to insure we are making a typical adjustment. To show what has happened, we use Image>Adjustments>Levels to display the histogram again.

This looks reasonable and so does the resulting picture, but this Levels graph nearly hides the fact that both highlight and shadow detail have been lost. To better see this, we will make a –10 contrast adjustment on the picture (Image>Adjustments>Brightness/Contrast) to pull the pure white and pure black at the ends of the histogram back in and use Levels again to show the result.

Note the spikes at both ends of the histogram. Instead of tailing off at either end as it did originally, the graph just stops with a spike peak at each end. That shows Auto Contrast has increased the contrast enough to have driven some grays at the dark end to pure black and some grays at the light end to pure white. Both shadow detail and highlight detail have been lost. Although there are exceptions, nearly every time you see a histogram plot of one of your pictures with vertical spikes at the end of the histogram, something has happened to make you lose highlight or shadow detail. The larger and sharper the spike, the more detail that has been lost.

But isn't this the same thing that happens in silver-based photography, where blocked highlights and loss of shadow detail are not uncommon? The short answer is no, because the behavior of the two are quite different. The reason has to do with the S-curves of silver-based photography, which produce a gradual loss of detail rather than a sharp cut-off. (See our white paper (CFS-211b) for a mathematical explanation.) Here we show how to use Photoshop to achieve an analogous gradual loss, using the Levels and Curves tools.

First, we revert to the original picture and use the Levels tool to adjust the contrast range by pulling the end points in until they just meet the tips of the histogram tails. This produces a similar histogram to the Auto Contrast adjustment but without spikes on the end:

Now we use the Image>Adjustments>Curves tool, making an adjustment like the following:

The unadjusted starting "curve" at left is a straight diagonal line. When we adjust this line, in the ranges of gray tones where it becomes steeper the contrast will increase and in the ranger of tones where it is less steep, contrast will decrease. Auto Contrast effectively makes an adjustment so the line remains straight, with the end points moved over, as shown in the middle curve. (The trim adjustment needs to be made before applying either the curve at the center or the one on the right. A straight line curve representing the complete Auto Contrast adjustment would be steeper, with the end moved even farther away from the diagonal corners.) All light grays in the original picture that were whiter than the light gray where the line touches down at the bottom will become pure white, and any dark grays darker than the dark gray where the line touches the top will be pure black. To produce the curve on the right, we clicked in three anchor points on the central part of the straight line, moved the end points back to pure white and pure black, and then gracefully curved the line at each end. By doing this, none of the gray tones have been converted to pure black or pure white. This effect is accepted by the eye as very natural – often much more so that the alternative loss of shadow and highlight detail, and in fact this is the way silver-based materials behave naturally.


The difference, even for this low quality photo, can be seen in the above detail comparison. The detail is from the lower left edge, the triangle of stone fence just above the shadow of the foot. Both images have been darkened using the Levels tool so the highlight detail shows more clearly. The picture on the left that we produced with the Curves tool shows the stone texture detail quite clearly while the picture on the right, produced using Auto Contrast, has completely lost detail through most of the triangle.

We've said that whenever you have a histogram plot that has vertical spikes sticking up on either end, you have lost highlight or shadow detail. While that is generally correct, some losses are critical to a picture while others may be unimportant. In fact, in other Photoshop tips we suggest as a rule of thumb trimming off about half of the nearly zero tail on the highlight side and all of the nearly zero tail on the shadow side. Like all such rules, it is often useful but sometimes wrong. However, when you see the warning spikes appear at either end of the histogram a simple test will tell if the detail loss will be a problem. First, convert the picture to 8 Bits/Channel mode. Then double-click on the Foreground color patch and set the foreground color to R=0, G=0, B=0, which is pure black, the color to be used for the following selection. From the menus, Select>Color Range and set Fuzziness to 0, and then OK. Now, with the selection active, Image>Adjustments>Invert. All of the formerly pure black pixels are white. Use the magnifier to explore and see if any of the areas are large and should have detail. Now use History to go back just prior to the Color Range step. Set the Foreground color to R=255, G=255, B=255, pure white, and repeat the selection, inversion, and exploration. All the formerly pure white pixels will show as pure black.

As you make these selections above, you can save them: Select>Save Selection, then for the Document pulldown select New, set the name, perhaps as Dark or Light, and then OK. When saving the second selection, from the Document pulldown select the Untitled image you created saving the first selection. Now use history to go back to a stage of the picture prior to when the spikes appeared, Select>Load Selection, pick the Untitled file from the pulldown and select Light or Dark, and you will be able to examine the detail that was lost. Using the Levels command with this selection set will help evaluate the potential of the lost detail to be useful.

If you try the above on the test picture for this tip, you will find the lost highlight detail in the triangle of stone as we have illustrated and you will also find that the lost shadow detail largely results from the paper texture of the print, and is just as well lost.

For preliminary, major manipulations of photographs such as we have done here is best to scan the image in 16-Bits/Channel mode and continue the major operation in that mode, converting to 8-Bits/Channel afterward to make tuning adjustments if necessary. If it is not possible to scan and work in 16-Bits/Channel the methods will still work, but are more likely to show stair-step effects. We do an adjustment very similar to the above on the majority of photographs we work with when we want to make large, high-quality prints. It is a matter of judgment and experience, watching the results on the display and sometimes trying several times to get precisely the wanted adjustment, with the initial levels trimming usually done separately for each channel of a color photograph, as we describe in Color Balancing Difficult Cases. We avoid the automatic adjustments except for quick proofing because they often do precisely the wrong thing, losing highlight and shadow detail as described above.

Return to Tips List