Innovations
in Applied Mathematics relating to Sight and Sound
Innovations
for Non-Profit Corporations relating to Financing and Housing
Go to C F Systems Home Page
Privacy Policy and E-Mail
Photoshop Tips for Advanced Users
Last Changed 10-24-14 but not revised since 2012
The ColorPerfect Photoshop/Photoline plug-in is available for many versions of Photoshop, Photoshop Elements, and Photoline for both PCs and Macs.
ColorPerfect is currently in version 2.15. (See all about ColorPerfect and download a demo version.)
The tips presented here are for the advanced Photoshop user working with photographic images. A number of them are still quite valid, while others are have been supplanted by better information. I have made notes to that effect in some of them, but this is low on my currrent list of priorities and I have not made any recent or systematic checks. So be aware that I no longer hold with some of the things said here long before my extensive studies into color imaging.
There is a 16-Bits/Channel Highlight Detail Secret tip with a surprising demonstration of 16-Bits/Channel power. If you have been having trouble with blown highlights in digital images this may be the secret you've been looking for.
Complete Color Integrity
ColorPerfect is the result of a surprising new discovery for handling an old problem: Several years ago in these web pages I began to explore the matter of color integrity in digital imaging. This came about because like most people, when I started into digital photography I was hypnotized by the marvelous things that are possible even easy. Then the marvel began to wear off when I started to realize that something was wrong with the color in many digital images. Not all digital images, but in so many of them; not only in snapshots, but in professional photographs, even in the highest quality magazines. Worst of all, I found this disturbing color effect in some of my own photos.
I did have success in finding the main cause of this lack of color integrity but there was a serious problem in that my early work was mathematical and difficult for most people to appreciate. Very recently I have had a real breakthrough that has so greatly expanded my understanding of color integrity that now I can explain it in simple terms and without using mathematics. In fact, after seeing this explanation you may find it difficult to believe that it is not already a core element of digital imaging. Yet, clearly it is not. The tools for maintaining color integrity prove to be very basic and very simple indeed, yet you will not readily find them in Photoshop. As this is being written popular tutorials and common practice both emphasize tools that actually destroy color integrity instead. That is why the problem is so pervasive even in professional work.
I feel very fortunate to have made this breakthrough. It is a real rarity to have several lines of study of a complex topic converge so beautifully into such a simple explanation. Join us on our Complete Color Integrity page to learn the full story − also available as a PDF document (1.25 MB PDF).
Tips
Our investigations into the problems of accurate digital color representation have led to several web pages on our site which we believe all serious photographers will find both revealing and helpful. The first and most central of these is a demonstration/explanation of the serious and pervasive problem of Color Integrity in Digital Images. Our Photo Documents web page covers this as one of our recent research projects into the principles of digital and film photography. For each project there is a non-mathematical, practical summary and an in depth mathematical analysis. The approach is largely original with us and we have not seen similar treatments elsewhere, either in print or on the web. These documents answer many of the mysteries of digital photography.
We have some comments about a novel method of sharpening photographic images, comments on precision and accuracy in digital photography, several answers to "Where are the color density filters?", a novel perspective on routine color balancing, a strange but effective way to rough color balance difficult cases, a method for coloring B&W photos, and avoiding loss of detail. There is a tip on color balancing color negatives that fully explains what is involved and why and another tip for B&W photographers who want to achieve an accurate tonal scale.
These tips are original with us and are not likely to be found elsewhere. You are welcome to use these tips in articles, courses, or books, but please credit C F Systems and www.c-f-systems.com when
you do. E-mail us (cfs@colorperfect.com) if you wish. Please see NOTICE in the paragraph below.
It
may be inspiring to know that these novel tips were developed by
a member of the older generation.
Having made my first color print at age fifteen, I have
over fifty years experience in printing color photographs as a
hobby. Adding to this is more than forty-five years experience dealing
with computers, mathematics, and numerical methods. This gave me a
unique perspective when I started using Photoshop several years
ago. The tips on this page will mostly avoid the mathematics, to
make them useful to more people. If you are curious about the origin
of a tip and you indicate your level of understanding in mathematics
when you e-mail us (cfs@colorperfect.com) , we will try to respond accordingly. We try to respond to all legitimate e-mails that we receive. If you do not get a response, please see the NOTICE on our Privacy Policy Page.
A
Better Method of Sharpening Pictures in Photoshop
The Simply Amazing Complete Color Integrity
A 16-Bits/Channel Highlight Detail Secret
Color Integrity in Digital Images
Color Negative FAQ
Color Density Filters in Photoshop
Routine Color Balancing?
Color Balancing Difficult Cases
Color Balancing Color Negatives
Getting a Proper Tonal Scale in Black and White Photography
Matching Colors in Photoshop
Preventing Banding and Stair-Step in Photoshop Images
Accuracy and Precision in Digital Photography
Coloring B&W Photos
Brightness, Contrast and Loss of Detail
Silver-Based Gamma, Video Gamma and Levels
Processing Old Color Films
Visit Our Photo Album
Although
it has nothing to do with Photoshop, our monograph on the problem of
highlight and shadow detail in digital photography may also be of
interest. Go to Highlight/Shadow Page
As
used in the context of this page, "Adobe" and "Photoshop"
are registered trademarks of Adobe Systems Incorporated.
A 16-Bits/Channel Highlight Detail Secret
This tip would have been on our site earlier but we only recently learned that very few photographers seem to be aware of it. The tip includes a neat demonstration of the power of 16-Bits/Channel images. The basic idea is simply that if you are scanning in 16-Bits/Channel images or bringing in 16-Bits/Channel RAW digital image originals and you are losing highlight detail, just darken the image. Make the image quite dark if necessary to bring out the highlight detail. For reasons we will demonstrate, you do not lose shadow detail by doing this and it is not at all necessary to separately scan a light and dark image and combine them. Once you have the dark image in Photoshop you can use various tools to selectively lighten portions of the image, preserving the highlight detail as you brighten the too-dark parts of the image. With Photoshop RAW input, darken the image not with the brightness control (which is not brightness at all) but with the exposure control, moving it into its negative value range. The exposure control also doctors the image somewhat but generally will be acceptable in this usage. (Note: How well this works with Photoshop RAW may depend upon the type of camera. Fuji cameras with the SuperCCD sensor chip are designed to capture extra highlight detail and work well. However, if the camera does not capture the highlight detail in the first place, this tip will not work.)
Scanner software is all different, of course, and the controls are often misleading, so telling which control will darken the image without otherwise altering it can be tricky. You can test whether you have chosen the correct darkening tool in the scanner by first scanning an image normally and then, without moving the image source, scanning a darkened version. Use a moderate resolution (dpi) well below the scanner's maximum to avoid confusion from scanner variability. If you have chosen the correct tool it should be possible to use just the highlight slider on the Levels tool to adjust the darkened version so that it exactly matches the normal version. For a really precise test first Image→Adjustment→Invert the normal image. Now select all of the darkened image, copy and paste it over the inverted normal version. Then, operating always on the upper, darkened image layer select Layers in the Layers-Channels-Paths dialog and set the opacity to 50%. Now, using the Levels tool and sliding the highlight adjustment pointer to the left it should be possible to achieve an evenly gray image in which the underlying image becomes invisible. After you OK out of the levels tool, explore the image with the cursor so that the pixel values show up in the Info panel. The pixel values should all be the same. Variations of two or three counts are acceptable. This test works better in 16-Bits/Channel mode, which requires Photoshop CS. You can use this same test with Photoshop RAW to produce normal and dark images and see how much the "exposure" control deviates from what it should do.
If you were to darken an 8-Bits/Channel image in an effort to preserve a highlight detail you would lose a lot of shadow detail. It is natural to think that the same will happen with a 16-Bits/Channel image but that is not at all the case. Photoshop displays pixel values as 0 through 255 for both 8-Bits/Channel and 16-Bits/Channel images so it is not generally appreciated that the space between any two successive counts in a 16-Bits/Channel image can hold a full 8-Bits/Channel picture. That is, a fully detailed color 8-Bits/Channel image could reside between pixel values of 100 and 101. To illustrate this we have a sample image in 16-Bits/Channel: Colorkid2.tif (400 KB) (To download this file using IE, "wrong" click the mouse and select "Save Target As." Other browsers have similar tricks.) If you load this image into Photoshop and explore it with the cursor so that the pixel values show up in the Info panel you will find that all pixel values are given as either 0 or 1; nothing higher. Now use Image→Adjustment→Levels and slide the highlights slider to the left as far as you can. Normally this will reduce the highlight value from 255 to 2 - it will not continue to 1 or 0. OK out of levels and you will have a color image, but quite dark. Now call Levels again and again bring the highlights slider to the left, but go only to 128 this time. OK out of Levels and you will have a full color image, all of which originally showed pixel values of either 0 or 1. If it would have been possible to put the highlight slider all the way to 1 the first time, the second adjustment would not have been necessary. This illustrates that the detail you are shoving down into darkness by making the original scan dark is not lost and is easily recoverable.
If you were curious enough to look at the final histogram of Colorkid1.tif you probably noticed that it is a "comb" with every alternate cell having a value of zero. This is because Photoshop 16-Bits/Channel is really 15-Bits/Channel.
Once you have the darkened image, there are numerous ways to restore highlight detail, perhaps involving Levels adjustment layers applied with a gradient, or feathered selection of light and dark areas with different Levels adjustments applied. We will mention one here that we have seen less often. First put the image into Lab mode: Image→Mode→Lab Color. Then Image→Adjustment→Curves. Apply a curve something like:
Where the line is straight it preserves tonal values and where it curves over at the top it compresses tonal values, preserving highlight detail rather than blowing highlights. Because we are in Lab mode color integrity will be reasonably well preserved even for more elaborate curves. This is not true for the Curves tool in RGB mode. Of course it is possible to convert back to RGB mode after making this adjustment.
Return to Tips List
Coloring B&W Photos
This method of coloring B&W photographs in Photoshop can produce very convincing results. It also offers a lot of control, so the artistic possibilities are unlimited. There are Photoshop plug-ins and other programs that automate this same task to varying degrees, but here we will use just Photoshop.
The principle involved is the same one that has been used for a century or more, where B&W photographs have been colored by painting over areas of the photo with transparent media, usually special transparent oil paints. A wash of a single flesh tone over a monochrome image of a face or of green over the leaves of trees or red over a plaid shirt is normally very convincing even though the coloring carries no fine detail. As a personal historical note, the original reason I took up darkroom work was to make mat-finish prints so my sister could paint them; that starting a couple of years before the example photograph used here was taken.
Coloring photographs can be done most satisfactorily in Lab mode. Lab mode does not refer to "laboratory," but to the three components of its color system, with "L" being lightness-darkness, "a" being a measure of red versus green, and "b" being a measure of blue versus yellow. This system is designed to be perceptually even, as nearly so as is practicable, with the result that major color and density changes made under the Lab system are much less prone to producing color casts and odd color effects than the familiar RGB system. It is well worth learning about Lab just because it often offers a ready solution to otherwise difficult Photoshop color problems. Here, the grayscale image we start with is a reasonable approximation to the "L" lightness scale, with the "a" and "b" components each starting out as zero, or no color. We will select an area and change "a" and "b" to apply a color cast over that area in exactly the same way as applying a transparent wash of oil color would work on the same area of a paper print.
The example shown here is a "Portrait of the Artist as a Young Man" which has been colored by the method to be described. This was not chosen for its artistic merit, but serves well as an example.
This photograph is from a paper negative made around 1953. At that time I took a piece of print paper intended for making enlargements and taped it where the film would go in an old folding camera. Enlarger paper is very slow compared to camera film. Using available light coming in from windows, an exposure of a minute or more was necessary. Note that the image of myself and the less noticeable cat are both slightly blurred from motion in the long exposure, while the background is sharp. Like the "orthochromatic" films used for almost all amateur B&W photography until the 1960s, the enlarger paper was not sensitive to red light. That point will have importance later.
To color a B&W photo, start with a picture that you have already adjusted to be a good grayscale image. For best results scan a 16 Bit/channel mode image and remain in that mode at least until the Lab coloring has been completed. If that is not possible, 8 Bit/Channel mode can be used and will usually be satisfactory. Convert the adjusted grayscale image to Lab: Image > Mode, select Lab Color. Next make selections of the different areas that each will be given a uniform color, using any convenient method. The selections do not need to be exceedingly precise, but they do need to outline areas of uniform color. Working with a grayscale image it is often easiest to use the lasso to select, cleaning up the selection with the lasso Add to Selection and Subtract from Selection. After a selection has been made, save it using Select>Save Selection. If you are working in 16-Bit/Channel, the selections will have to be saved in a separate (New) image file.
While the area outlines do not have to be extremely precise, the different areas do need to meet at precise borders so that there are no gaps or overlaps. With the example photograph, I started by outlining the entire person and saving that selection as Person. For this process to work best, do not apply any feathering to the selections as you make them and save them not yet. Next, with the Person selection loaded, I used the lasso subtract to remove everything below the head and then more carefully to remove everything below the hairline. The resulting selection was saved as Hair. Then Person was reloaded (Select>Load Selection), followed by reloading Hair with "Subtract from Selection" selected. This insures that the upper border of the next selection will exactly meet with Hair. As with the hair, the face was selected by lasso subtracting what remained below the face, and the hands were lasso-added to the selection, the resulting selection saved as Flesh. Next, Person was reloaded, and Flesh was reloaded with "Subtract from Selection." The result was lasso-subtract trimmed below and saved as Shirt. And so on. Later, the entire wicker sofa was outlined, the outline passing through the seated subject, and then Person was reloaded with "Subtract from Selection" on. The entire set of books was outlined and then groups of books selected out of that selection. The overall books selection was reloaded and the already selected groups of books subtracted out prior to making each successive selection of books. These groups of books were used to get different colors of books on the shelf.
In all the above, lasso select was used because I have found working with that quickest and easiest for me. By all means, use other selection options if you prefer. Of course, some powerful selection options do not work well with grayscale images, regrettable but unavoidable. Many powerful selection options and techniques do not work in 16 Bit/Channel mode, also regrettable, but in this case easily solvable. Change the mode to 8 Bits/Channel and save the picture as a separate file. Make and save all the selections in this file, using any of the methods available in 8 Bits/Channel mode, but avoid cropping or resizing the picture. When the selections have been made and the new file saved again, leave this file open and re-open the 16 Bits/Channel file. Now you can load the selections while editing the 16 Bits/Channel image by using Select>Load Selection and picking the name of the 8 Bits/Channel file from the upper pulldown list.
The example image is available in PSD form, as a reduced 800 x 499 Lab image (the original working version was 2001 x 3205) with each selection represented as a levels adjustment layer. This shows how the selections are arranged and what Lab adjustments were made to each selection, to give an idea of what to do with your own picture. The file will work only with Photoshop and is a little more than a megabyte in size. (Download Example Image.)
Working with your own picture file, next load the selections one by one and apply color. It is usually better to add some feathering perhaps 3 or 4 pixels for a 2000 x 3000 photo at this point unless the borders of the selection are truly sharp. To apply color, use Image>Adjustments>Levels. Since you are in Lab mode, the selections are "Lightness", "a", and "b" rather than "RGB", "R", "G", and "B". Select the "a" channel. Note that the profile is a vertical line at a=0, or very nearly so. Move the center slider to the left and the image will become more red to the right more green. Try to pick a reasonable tone for the area this will become easier with experience. Next, do the same with the "b" channel. Move the center slider to the left and the image will become more yellow to the right more blue. Again, try to pick a reasonable tone. Alternately adjust a and b until you think you are close. Do this with each area selection until the photograph is completely colored. Then go back and readjust the areas that need improvement.
Particularly when you are first starting into this, it will be helpful to have a reference. Pick a region in the example photo and see what was done to achieve that color, then try that as a starting point for a similar area in your own picture. It may also be helpful to load an actual color picture of similar content into Photoshop and convert it to Lab mode. Then select that picture and hover the pointer over an area so the Lab coordinates will show up in the Info window (Window> check Info). Select the picture you wish to color and hover the pointer over a similar area to see its Lab coordinates. Alternately adjust the Levels for the selected area and hover the pointer over it until you approximately match the a and b coordinates to the reference picture.
In the above, we did not adjust Lightness. That seems reasonable because a and b are the color channels. However, Lightness really is an integral part of the color system and you will find that sometimes Lightness will have to be adjusted a little to get the color you want. Again, use the center slider to adjust lightness, then go back and tweak a and b. Before, we mentioned that older B&W photographs are likely to be orthochromatic, as is our example. For such images the Lightness is likely to be quite far off for some colors and may need fairly large adjustments to look reasonable. So, do not be afraid to adjust Lightness. In fact, Lab mode is the best place to make such overall density adjustments when they are necessary. In the example file, you will find that the Blanket and Wall Paper have curves adjustment layers in addition to the levels adjustment. For each of these I wanted the lighter portions to fade to white; the curves will show the strange adjustment necessary to achieve this while maintaining a consistent overall color.
When the photograph looks like it is close to correct, convert back to RGB mode and if you wish, make final adjustments in a mode which is probably more familiar. If you are in 16 Bits/Channel mode, you may wish to change to 8 Bits/Channel mode. Selections that were saved in a separate file will still work in 8 Bits/Channel mode and long as the image is not cropped or resized. Particularly the first few times you try this method, you may want to save your Lab mode working file and revisit it after a few days; very likely some areas that appeared OK previously will appear less so after the eyes have rested. It is particularly easy to misjudge the saturation of colors. Large errors in saturation are best handled in Lab mode, but if saturation errors are minor, you may find it easier to correct them in RGB mode, using the Hue-Saturation control.
It is likely that you can make significant improvements in the photograph using RGB mode, due not only the experience factor but also because any switch of mode always will make some adjustments easier. For example, by using Image>Adjustments>Curves to edit a colored selection, the endpoints of the R, G, and B curves can be adjusted to purposely add or remove color casts in the lighter or the darker parts of the selection. The PSD example file was tweaked in RGB mode to form the JPG version that appears above.
Return to Tips List
Getting a Proper Tonal Scale in Black and White Photography
Photoshop does just as poor a job of inverting black and white negatives as it does with color negatives, the resulting tonal scale being extremely warped. While the ColorNeg plug-in was designed to work with color negatives, it does black and white negatives equally well, giving complete control over the image gamma in the process. By doing the inversion properly, ColorNeg produces an accurate tonal scale and in the process, provides the tools to manipulate the tonal scale in a manner very similar to what Ansel Adams did in the darkroom. The method of dealing with the tonal scale described in Color Balancing Color Negatives can be used in place of ColorNeg, although it is considerably more cumbersome to do so.
Return to Tips List
Silver-Based Gamma, Video Gamma and Levels
Photographers with much experience in silver-based methods know that gamma is a measure of the contrast shown by a photographic material. Many people know that the middle gray Levels tool in Photoshop really is a "gamma adjustment." Is this the same gamma that indicates contrast? I have not found much written about this and apparently the reason is that the answer is complicated and can be confusing. The practical answer is no, the two gammas are really quite different even though the mathematical form is nearly identical. Download the mathematical PDF (175KB) for a complete explanation. This document has been completely rewritten with more detail and we believe it will be easier to understand [2/1/4].
Although the two equations for gamma in silver-based photography and video work appear nearly the same, in practice the behavior and meaning of gamma is quite different in the two fields. A higher silver-based gamma always means increased image contrast and a lower silver-based gamma always means lower image contrast. This is never the case with video gamma, which is what is actually used by the "Levels" tool in Photoshop. In Levels, moving the midpoint to the right makes the overall image darker, but this is achieved by lowering the contrast in the darker tones of the picture while simultaneously raising the contrast in the lighter tones of the picture. This is easily seen once you know what to look for. Similarly, moving the Levels midpoint to the left makes the overall image lighter, and this is achieved by raising the contrast in the darker tones of the picture while simultaneously lowering the contrast in the lighter tones of the picture.
Despite the fact that gamma is a fairly straightforward mathematical concept, there is considerable confusion regarding gamma in digital photography. It can be rather fun to track down the exchanges between Timo Autiokari and Charles Poynton concerning gamma (try Google if you are interested). These are two obviously intelligent people. Both offer very useful information on the web, but each clearly believes his view is the only one possible and the other fellow hasn't a clue. But there are very few things in science and virtually none in technology that can't be viewed in more than one valid way.
Return to Tips List