color balancing techniques

8/3/2019 Color Balancing Techniques

1/13

Color Balancing Techniques 1

Written by Jonathan SachsCopyright 1996-1999 Digital Light & Color

Color Balancing Techniques

Introduction

Color balancing refers to the process of removing an overall color bias from animage. For example, if an image appears too red, it is said to have a red cast.Removing this red cast brings the image back into balance. Color casts can arisefrom many different causes: the way the original scene was illuminated, the filmand filters you use, variations in film processing and printing, or from the scanningprocess. Since it is difficult or impossible to control all the factors that can create acolor imbalance, it is usually easiest to correct the problem at the end of the pro-

cess.


2/13


2 Color Balancing Techniques

Color correction using CC filters with theFilter Transformation

The traditional method of controlling color balance when printing color images inthe darkroom is to use one or more color correcting (CC) filters. CC Filters are agraduated set of standard color filters available in each of the primary and second-ary colors (red, green, blue, cyan, magenta, yellow). The filters are designated bytheir density and color. For example, a CC10M filter is a magenta filter with a den-

sity of 0.10. Filters with higher density transmit less light and are therefore darkerand more intensely colored. Color balancing an image using CC filters has twosteps. First you must identify the color cast of the image, and this takes a bit of practice. Then, to remove that color from the image, you must apply a filter whosecolor is the complement of the color you want to remove. For example, to removethe magenta cast in the image above, you would need to use a green filter (sincegreen is the complement of magenta). In addition, the density of the CC filter youuse must be proportional to the strength of the color cast in the original image.

This method of color balancing is fully supported by Picture Window with the Fil-ter transformation and setting the filter color using the color pickers filter bank option. You can select any combination of standard CC filters and adjust for the fil-ter factor using the Exposure Compensation slider. The procedure is as follows:

1. Open the input image and click on its window to select it.2. Select Transformation/Filter from the main menu. The Filter dialog box is dis-

played.

3. Click on the large square button just to the right of the caption Filter: in the dia-log box. Then select Solid Color from the menu that pops up. A color picker is

displayed.


3/13


Color correction using the Color Balance Transformation

4. Click on the button in the upper left corner of the color picker dialog box.Then select Filter Bank from the menu that pops up. This replaces the standardcolor picker color wheel with a set of colored buttons. Each button correspondsto a standard CC or Wratten filter and they are labeled with their filter designa-tion. If necessary, you can use the scroll bar to bring more buttons into view.

Each time you click on a filter button, the corresponding filter color is selected.You can combine as many different filters as you like. To deselect a filter button,

simply click on it a second time. To deselect all the filter buttons, use the Reset button. If a filter button is selected, it is displayed as depressed.

5. Click the Preview button in the Filter dialog box to see the effect of applying thefilter, and adjust the filter color as necessary until the desired result is obtained.If you select the Auto preview check box at the top of the Filter dialog box, theneach time you click on a new point in the color wheel, Picture Window willautomatically rebalance the image and display a preview of the result for you toevaluate. Comparing the corrected version with the original by arranging theinput and preview windows side by side is usually very helpful.If a filter makes the image too dark, you can adjust the Exposure Compensationslider to lighten the image. To select the complement of the color you have spec-ified, click the Complement check box. For example, to remove magenta froman image, you could select a magenta filter and then click the Complementcheck box to create a green filter that reduces magenta in the input image.

Color correction using the Color BalanceTransformation

In addition to the Filter technique for color balancing, Picture Window also has aColor Balance transformation created especially to make color balancing easy, evenfor those unfamiliar with the use of color correction filters. In addition to requiringless experience and skill, the Color Balance transformation can handle difficultcolor balance problems that simple filtration cannot correct.

The Color Balance transformation lets you add or remove color casts from animage's highlights, midtones, and shadows, and it can perform any combination of these operations simultaneously.

A perfectly color balanced image renders the full range of neutral grays, includingblack and white, without any tinge of color. Sometimes this is desirable, and some-


4/13



times it is not. For example, photographers often take pictures in early morning orevening light because of the dramatic shadows and the yellow-orange color of thelight that imparts a warm glow to everything it illuminates. One way to use theColor Balance transformation is to start by removing any color cast the input imagecontains, thus balancing it perfectly -- and then adding a deliberate color cast if thesubject matter calls for it.

Highlight balancing

The most visible color casts are those that affect the highlights. For example, if youview an image through a red filter, the darkest parts of the image will still look moreor less the same, but the lighter parts, especially pure whites, will take on a very dis-tinct red coloration. Adjusting the highlight balance of an image with the ColorBalance transformation produces an effect very similar to using the Filter transfor-mation, but you control the transformation differently.

The color balance dialog box contains two color patches for highlights -- one is the

color to remove and the other is the color to add. For example, if an image has a redcast you want to remove, just click on the remove highlight patch and use the colorpicker to set the color to red. The more saturated the red you select (the farther fromthe center of the color wheel), the more red is removed from the image. If youselect the Auto preview check box at the top of the Color Balance dialog box, theneach time you click on a new point in the color wheel, Picture Window will auto-matically rebalance the image and display a preview of the result for you to evalu-ate. Comparing the corrected version with the original by arranging the input and

preview windows side by side is usually very helpful.


5/13


Color correction using the Color Balance Transformation

In addition to choosing the color to remove from the color wheel, there are twoother methods that are sometimes much simpler to use: automatic highlight castremoval and setting the highlight color to remove by using the highlight probe onthe input image.

Auto color balancing

When the color balance transformation starts up, it make a pass over the entire

image and attempts to determine the color of the very brightest and darkest parts of the image. If you select Automatic highlight or shadow color balancing, the high-light or shadow color to remove is set to the color Picture Window has found in thisinitial analysis. Sometimes this automatic color balancing works perfectly. If so,then you're done. If not, you will have to set the color to remove manually or by theprobe method (see below).

Selecting the color to remove using the probe

To activate the probe, you use the Probe control in the Color Balance transforma-tion dialog box. To perform highlight color balancing using the probe, you set theProbe control to Highlight. At this point, when you click on some part of the inputimage, the color to remove is set to that color and the color to add is set to a neutralgray of the same brightness level. Thus if there is some part of the image that youknow should be rendered as a neutral gray, clicking on it with the highlight probewill automatically compute the color to remove so as to neutralize the selected

color. This is often the quickest and most effective method for removing a colorcast. For portraits, you can use the whites of the eyes as a reference neutral.

You can vary the probe size by clicking on the button in the Color Balance dia-log box. A small probe is useful for selecting a very small patch of color; a largerprobe works better when sampling part of an image that is somewhat irregular anddifficult to get an average reading from.

In addition to adjusting the color of the highlights, you can also adjust their bright-ness level (see Adjusting Dynamic Range below).

Midtone balancing

Midtone balancing is a technique that can work wonders for certain images. Forsome reason, especially with images scanned from color negative film, a situationarises where an image's highlights are a more of less neutral white and the shadowareas are a more or less neutral black or dark gray, but the midtones have a distinct


6/13



color cast. This is a color balance problem that cannot be corrected using a simplefilter because you can only correct the midtones at the expense of creating animbalance in the highlights.

The Color Balance transformation lets you select a midtone color to remove eitherby using a midtone probe or by selecting a color directly using the color picker. If the subject contains an object in the midtones that should be more or less neutral,the midtone probe is often the fastest way to remove the color cast.

Shadow Balancing

Sometimes the shadow areas of an image are not a neutral black. You can correctthis problem by setting the shadow color balance. In addition to adjusting the colorof the shadow areas, you can also adjust their brightness level (see AdjustingDynamic Range below).

Interpreting the color curvesThe color balance transformation works by applying three curves to the inputimageone for the red channel, one for green channel, and one for the blue chan-nel. These curves are displayed on the right side of the dialog box.

Each curve is defined by three points corresponding to the current highlight, mid-

tone, and shadow color selections. You can either set the curves by choosing colorcasts to add and remove, or you can drag points on the curve to select the colors.

Correcting for fluorescent lighting

Fluorescent lights often introduce a greenish cast to photographs. Removing thiscast, especially from scans of color negatives, often requires use of a midtone castas well as a highlight cast. Set your highlight cast first to get the image highlights


7/13


The Color Correction transformation

the right color, then work on the midtones. In an image with mixed fluorescent andother lighting, you may have to create a mask to isolate the areas illuminated by dif-ferent light sources and balance them separately.

Adjusting dynamic range

To adjust the dynamic range of an image, first set highlight and shadow to auto andbring up color pickers for the highlight and shadow casts to add by clicking on their

respective color patches. To increase the dynamic range, move the value slider of the highlight color picker up and the slider of the shadow color picker down. Thismakes the highlights brighter and the shadows darker. Moving the sliders in theopposite directions reduces the dynamic range of the image.

Applying consistent color balance changes to several images

Adjust the first image and then use the Save As option (via the button) to savethe color balance settings in a file, then you can reload the settings for use withother images.

Color matching to a reference image

If you want, for example, to match a skin tone from one image to a skin tone inanother image, all you have to do is select the skin tone in the input image as the

highlight cast to remove and the skin tone you want to change to as the highlightcast to add. The procedure for doing this is as follows:

1. Set shadow and midtone balancing to none .2. Select the highlight probe and click on the input image to select the skin

tone as the highlight cast to be removed.3. Click on the color square for highlight cast to add; this pops up a color

picker. Click on the eyedropper icon in the color picker.

4. Click on the skin tone you want to match; this sets the highlight cast to add.


Both Filter and the Color Balance apply corrections uniformly to the entire image.In many cases, this works well, but sometimes the situation is a little more compli-cated. For example, Fuji Velvia has a tendency to produce highly saturated reds


8/13



while reproducing other colors more normally. While this works wonders in the redrock country of Utah, it can wreak havoc on skin tones or even pine needles. If youtry to tone back the reds, you will then start boosting the greens and blues in theimage as well (see below). This phenomenon is sometimes referred to as crossover and cannot be corrected with a color correcting filter.

While some problems of this type can be fixed using the midtone balancing featureof the Color Balance transformation, the Color Correction transformation (availableonly in the Pro version of Picture Window) is the ultimate tool for making just thiskind of selective color adjustment. It works by creating custom color lookup tablesthat apply different corrections to different parts of the color space. For example,this lets you adjust just the reds in the sample image while leaving all other colorsin the image unchanged.

Original Filtered to reduce red

Original Reds reduced using Color Correction


9/13



The Color Correction transformation is based on the same color wheel Picture Win-dow uses in its color picker. The color wheel is a hexagon with white in the centerand the fully saturated colors around its perimeter. Every possible combination of hue and saturation in the HSV color space corresponds to a different point on thecolor wheel. The higher the saturation, the further the point is from the center; vari-ations in hue corresponding to rotations about the center. All the colors in the colorwheel correspond to the maximum possible brightness (V = 100%); every possiblecolor can either be found somewhere on the color wheel or can be generated by

darkening one of the colors on the color wheel by reducing each of its componentsby the same fraction.

To specify which colors are preserved and which are changed, the Color Correctiontransformation uses control points . Each control point consists of two parts: aninput color and an output color. When the transformation is applied, the input imageis examined one pixel at a time. If the hue and saturation of the pixel (regardless of its value or brightness) matches the input color of any of the control points, thepixel will be changed to the output color with the same brightness in the resultimage. For colors between control points, the output color is interpolated using thenearest available control points so there are no abrupt transitions between colors.

To preserve colors in one region of the color wheel, you create a control pointwhose input and output colors are identical. To make one color (and those in itsimmediate neighborhood) change to another, you create a control point whose inputcolor is the one you want to change and whose output color is to one you want tochange it to.


10/13



Control points are displayed on the color wheel. If the input and output colors of acontrol point are the same, it is displayed as a small circle at the correspondinglocation on the color wheel. If the output color differs from the input color, then anarrow is extended from the input color to the output color, indicating that the inputcolor will turn into the output color.

Initially, there are seven control points, each of which have the same input and out-put color. One is located at the center of the color wheel (white) and one at eachvertex of the hexagon (red, yellow, green, cyan, blue, and magenta). Nailing downthe central white point guarantees that white (and consequently all neutral grays)will be left unchanged. Similarly, nailing down the fully saturated primary and sec-ondary colors at the vertices of the hexagon keeps them from being changed. Run-ning the transformation with these settings will, however, leave the input imageunchanged as all the output colors are the same as the input colors.

While you can change either the input or output colors of any of these initial sevencontrol points, you can also add new control points or remove existing ones. Thereare two ways to add a new control point:

1) You can shift-click on the color wheel to add a control point at a specificlocation.

2) You can enable the Probe (by depressing the eyedropper icon just to the rightof the color wheel) and then click on a point in the input image window thathas a color you want to change or preserve. This adds a new control point at thelocation on the color wheel corresponding to the color of the image at the pointyou clicked.

To change the output color of a control point, just click and drag it -- this causes anarrow to grow out of the control point whose tail is the input color and whose headis the output color. Once you separate the two, you can change the input color bydragging the little circle or the output color by dragging the head of the arrow.

To remove a control point, control-click on it.


11/13



As an illustration, lets see how to use this transformation to change reds in animage without changing anything else.

Leaving the initial seven control points in place, one additional control point wasadded by simply clicking on the eyedropper icon and then clicking on a sample of the red pine needles in the input image window. The control point was then draggedtowards the center of the color wheel (see the arrow above) to more or less halve itssaturation while leaving its hue the same, yielding the much more natural effect at

the right. Note that the foliage, sky, and rocks have not been changed at all, nor havethe whites in the image been contaminated. This kind of very selective color bal-ancing is very difficult to perform any other way.


12/13



In the following example, the more or less neutral waterfall can be made more bluewithout deadening the yellows, reds, and greens in the foliage by dragging just thewhite point:.

If the same color shift was introduced into the image using the Filter or Color Bal-ance transformation, the results would look like this:.

Note the loss of vibrance in the foliage above the waterfall.

The Color Correction transformation can also be used with color printers. Someday, precision color management will be seamlessly integrated into all Windows

printer drivers and images will all look the same on paper as they did on the moni-


13/13


Use Masks for Even More Control

tor. Until then, the situation is less ideal. If your printer consistently renders certaincolors incorrectly, you can use the Color Correction transformation to create (andsave for later reuse using the Opt button) a set of lookup tables that compensates foryour printers idiosyncrasies. If a particular color is coming out with the wrong sat-uration or hue, simply create a control point and drag the color in the oppositedirection. With some trial and error, you should be able to correct at least the moreobvious problems.

Use Masks for Even More Control

For an extra level of control, the Filter, Color Balance, or Color Correction transfor-mations can be used with a mask.

If only one section of an image needs correction, you can create a mask that isolates

just that part of the image and restrict the color correction to that region.

Density masks are useful when you want to make color corrections just to the shad-ows, midtones, or highlights of an image. A density mask is simply a mask whosetransparency at any given point is determined by the brightness of the image at thatpoint.

For more information on creating and using masks, please see the Picture Window

manual, the help file, or the white paper on Creating and Using Masks.

color balancing techniques

Documents