Table of Contents
One of the most common uses of the GIMP is to fix digital camera images that for some reason are less than perfect. Maybe the image is overexposed or underexposed; maybe rotated a bit; maybe out of focus: these are all common problems for which GIMP has good tools. The purpose of this chapter is to give you an overview of those tools and the situations in which they are useful. You will not find detailed tutorials here: in most cases it is easier to learn how to use the tools by experimenting with them than by reading about them. (Also, each tool is described more thoroughly in the Help section devoted to it.) You will also not find anything in this chapter about the multitude of "special effects" that you can apply to an image using GIMP. You should be familiar with basic GIMP concepts before reading this chapter, but you certainly don't need to be an expertΓÇôif you are, you probably know most of this anyway. And don't hesitate to experiment: GIMP's powerful "undo" system allows you to recover from almost any mistake with a simple Ctrl-Z.
Most commonly the things that you want to do to clean up an imperfect photo are of four types: improving the composition; improving the colors; improving the sharpness; and removing artifacts or other undesirable elements of the image.
It is easy, when taking a picture, to hold the camera not quite perfectly vertical, resulting in a picture where things are tilted at an angle. In GIMP, the way to fix this is to use the Rotate tool. Activate this by clicking its icon in the Toolbox, or by pressing the ΓÇ£RΓÇ¥ key capitalized) while inside the image. Make sure the Tool Options are visible, and at the top, make sure for ΓÇ£Affect:ΓÇ¥ that the left button (ΓÇ£Transform LayerΓÇ¥) is selected. If you then click the mouse inside the image and drag it, you will see a grid appear that rotates as you drag. When the grid looks right, click or press the enter key, and the image will be rotated.
Now as a matter of fact, it isn't so easy to get things right by this method: you often find that things are better but not quite perfect. One solution is to rotate a bit more, but there is a disadvantage to that approach. Each time you rotate an image, because the rotated pixels don't line up precisely with the original pixels, the image inevitably gets blurred a little bit. For a single rotation, the amount of blurring is quite small, but two rotations cause twice as much blurring as one, and there is no reason to blur things more than you have to. A better alternative is to undo the rotation and then do another, adjusting the angle.
Fortunately, GIMP provides another way of doing it that is considerably easier to use: in the Rotate Tool Options, for the Transform Direction you can select "Backward (Corrective)". When you do this, instead of rotating the grid to compensate for the error, you can rotate it to line up with the error. If this seems confusing, try it and you will see that it is quite straightforward.
Note | |
---|---|
Note: New in GIMP 2.2 is the option to preview the results of transformations, instead of just seeing a grid. This makes it easier to get things right on the first try. |
After you have rotated an image, there will be unpleasant triangular "holes" at the corners. One way to fix them is to create a background that fills the holes with some unobtrusive or neutral color, but usually a better solution is to crop the image. The greater the rotation, the more cropping is required, so it is best to get the camera aligned as well as possible when you take the picture in the first place.
When you take a picture with a digital camera, you have some control over what gets included in the image but often not as much as you would like: the result is images that could benefit from trimming. Beyond this, it is often possible to enhance the impact of an image by trimming it so that the most important elements are placed at key points. A rule of thumb, not always to be followed but good to keep in mind, is the ΓÇ£rule of thirdsΓÇ¥, which says that maximum impact is obtained by placing the center of interest one-third of the way across the image, both widthwise and heightwise.
To crop an image, activate the Crop tool in the Toolbox, or by pressing the ΓÇ£CΓÇ¥ key (capitalized) while inside the image. With the tool active, clicking and dragging in the image will sweep out a crop rectangle. It will also pop up a dialog that allows you to adjust the dimensions of the crop region if they aren't quite right. When everything is perfect, hit the button in the dialog.
In spite of sophisticated exposure-control systems, pictures taken with digital cameras often come out over- or under-exposed, or with color casts due to imperfections in lighting. GIMP gives you a variety of tools to correct colors in an image, ranging to automated tools that run with a simple button-click to highly sophisticated tools that give you many parameters of control. We will start with the simplest first.
GIMP gives you five automated color correction tools. Unfortunately they don't usually give you quite the results you are looking for, but they only take a moment to try out, and if nothing else they often give you an idea of some of the possibilities inherent in the image. Except for "Auto Levels", you can find them in the Layer menu, by following the menu path
→ → in the image menu.Here they are, with a few words about each:
This tool (it is really a plug-in) is useful for underexposed images: it adjusts the whole image uniformly until the brightest point is right at the saturation limit, and the darkest point is black. The downside is that the amount of brightening is determined entirely by the lightest and darkest points in the image, so even one single white pixel and/or one single black pixel will make normalization ineffective.
This is a very powerful adjustment that tries to spread the colors in the image evenly across the range of possible intensities. In some cases the effect is amazing, bringing out contrasts that are very difficult to get in any other way; but more commonly, it just makes the image look weird. Oh well, it only takes a moment to try.
Help me, what exactly does this do? Obviously it makes some things more saturated.
This is like ΓÇ£NormalizeΓÇ¥, except that it operates on the red, green, and blue channels independently. It often has the useful effect of reducing color casts.
This is done by activating the Levels tool (
→ → in the image menu), clicking on the image to bring up the tool dialog, and then pressing the button near the center of the dialog. You will see a preview of the result; you must press for it to take effect. Pressing instead will cause your image to revert to its previous state.If you can find a point in the image that ought to be perfect white, and a second point that ought to be perfect black, then you can use the Levels tool to do a semi-automatic adjustment that will often do a good job of fixing both brightness and colors throughout the image. First, bring up the Levels tool as previously described. Now, look down near the bottom of the Layers dialog for three buttons with symbols on them that look like eye-droppers (at least, that is what they are supposed to look like). The one on the left, if you mouse over it, shows its function to be “Pick Black Point”. Click on this, then click on a point in the image that ought to be black–really truly perfectly black, not just sort of dark–and watch the image change. Next, click on the rightmost of the three buttons ( “Pick White Point” ), and then click a point in the image that ought to be white, and once more watch the image change. If you are happy with the result, click the button otherwise .
Those are the automated color adjustments: if you find that none of them quite does the job for you, it is time to try one of the interactive color tools. All of these, except one, can be accessed via Tools->Color Tools in the image menu. After you select a color tool, click on the image (anywhere) to activate it and bring up its dialog.
The simplest tool to use is the Brightness/Contrast tool. It is also the least powerful, but in many cases it does everything you need. This tool is often useful for images that are overexposed or underexposed; it is not useful for correcting color casts. The tool gives you two sliders to adjust, for ΓÇ£BrightnessΓÇ¥ and ΓÇ£ContrastΓÇ¥. If you have the option ΓÇ£PreviewΓÇ¥ checked (and almost certainly you should),you will see any adjustments you make reflected in the image. When you are happy with the results, press and they will take effect. If you can't get results that you are happy with, press and the image will revert to its previous state.
A more sophisticated, and only slightly more difficult, way of correcting exposure problems is to use the Levels tool. The dialog for this tool looks very complicated, but for the basic usage we have in mind here, the only part you need to deal with is the ΓÇ£Input LevelsΓÇ¥ area, specifically the three triangular sliders that appear below the histogram. We refer you to the Levels Tool Help for instructions; but actually the easiest way to learn how to use it is to experiment by moving the three sliders around, and watching how the image is affected. (Make sure that ΓÇ£PreviewΓÇ¥ is checked at the bottom of the dialog.)
A very powerful way of correcting exposure problems is to use the Curves tool. This tool allows you to click and drag control points on a curve, in order to create a function mapping input brightness levels to output brightness levels. The Curves tool can replicate any effect you can achieve with Brightness/Contrast or the Levels tool, so it is more powerful than either of them. Once again, we refer you to the Curves Tool Help for detailed instructions, but the easiest way to learn how to use it is by experimenting.
The most powerful approach to adjusting brightness and contrast across an image, for more expert GIMP users, is to create a new layer above the one you are working on, and then in the Layers dialog set the Mode for the upper layer to ΓÇ£MultiplyΓÇ¥. The new layer then serves as a ΓÇ£gain controlΓÇ¥ layer for the layer below it, with white yielding maximum gain and black yielding a gain of zero. Thus, by painting on the new layer, you can selectively adjust the gain for each area of the image, giving you very fine control. You should try to paint only with smooth gradients, because sudden changes in gain will give rise to spurious edges in the result. Paint only using shades of gray, not colors, unless you want to produce color shifts in the image.
Actually, ΓÇ£MultiplyΓÇ¥ is not the only mode that is useful for gain control. In fact, ΓÇ£MultiplyΓÇ¥ mode can only darken parts of an image, never lighten them, so it is only useful where some parts of an image are overexposed. Using ΓÇ£DivideΓÇ¥ mode has the opposite effect: it can brighten areas of an image but not darken them. Here is a trick that is often useful for bringing out the maximum amount of detail across all areas of an image:
Duplicate the layer (producing a new layer above it).
Desaturate the new layer.
Apply a Gaussian blur to the result, with a large radius (100 or more).
Set Mode in the Layers dialog to Divide.
Control the amount of correction by adjusting opacity in the Layers dialog, or by using Brightness/Contrast, Levels, or Curves tools on the new layer.
When you are happy with the result, you can use
to combine the control layer and the original layer into a single layer.In addition to ΓÇ£MultiplyΓÇ¥ and ΓÇ£DivideΓÇ¥, you may every so often get useful effects with other layer combination modes, such as ΓÇ£DodgeΓÇ¥, ΓÇ£BurnΓÇ¥, or ΓÇ£Soft LightΓÇ¥. It is all too easy, though, once you start playing with these things, to look away from the computer for a moment and suddenly find that you have just spent an hour twiddling parameters. Be warned: the more options you have, the harder it is to make a decision.
In our experience, if your image has a color cast---too much red, too much blue, etc---the easiest way to correct it is to use the Levels tool, adjusting levels individually on the red, green, and blue channels. If this doesn't work for you, it might be worth your while to try the Color Balance tool or the Curves tool, but these are much more difficult to use effectively. (They are very good for creating certain types of special effects, though.)
Sometimes it is hard to tell whether you have adjusted colors adequately. A good, objective technique is to find a point in the image that you know should be either white or a shade of gray. Activate the Color Picker tool (the eyedropper symbol in the Toolbox), and click on the aforesaid point: this brings up the Color Picker dialog. If the colors are correctly adjusted, then the red, green, and blue components of the reported color should all be equal; if not, then you should see what sort of adjustment you need to make. This technique, when well used, allows even color-blind people to color-correct an image.
If your image is washed out---which can easily happen when you take pictures in bright light---try the Hue/Saturation tool, which gives you three sliders to manipulate, for Hue, Lightness, and Saturation. Raising the saturation will probably make the image look better. In same cases it is useful to adjust the lightness at the same time. ( ΓÇ£LightnessΓÇ¥ here is similar to ΓÇ£BrightnessΓÇ¥ in the Brightness/Contrast tool, except that they are formed from different combinations of the red, green, and blue channels.) The Hue/Saturation tool gives you the option of adjusting restricted subranges of colors (using the buttons at the top of the dialog), but if you want to get natural-looking colors, in most cases you should avoid doing this.
Tip | |
---|---|
Even if an image does not seemed washed out, often you can increase its impact by pushing up the saturation a bit. Veterans of the film era sometimes call this trick ΓÇ£FujifyingΓÇ¥, after Fujichrome film, which is notorious for producing highly saturated prints. |
When you take pictures in low light conditions, in some cases you have the opposite problem: too much saturation. In this case too the Hue/Saturation tool is a good one to use, only by reducing the saturation instead of increasing it.
If the focus on the camera is not set perfectly, or the camera is moving when the picture is taken, the result is a blurred image. If there is a lot of blurring, you probably won't be able to do much about it with any technique, but if there is only a moderate amount, you should be able to improve the image.
The most generally useful technique for sharpening a fuzzy image is called the Unsharp Mask. In spite of the rather confusing name, which derives from its origins as a technique used by film developers, its result is to make the image sharper, not ΓÇ£unsharpΓÇ¥. It is a plug-in, and you can access it as Filters->Enhance->Unsharp Mask in the image menu. There are two parameters, ΓÇ£RadiusΓÇ¥ and ΓÇ£AmountΓÇ¥. The default values often work pretty well, so you should try them first. Increasing either the radius or the amount increases the strength of the effect. Don't get carried away, though: if you make the unsharp mask too strong, it will amplify noise in the image and also give rise to visible artifacts where there are sharp edges.
Tip | |
---|---|
Sometimes using Unsharp Mask can cause color distortion where there are strong contrasts in an image. When this happens, you can often get better results by decomposing the image into separate Hue-Saturation-Value (HSV) layers, and running Unsharp Mask on the Value layer only, then recomposing. This works because the human eye has much finer resolution for brightness than for color. See the sections on Decompose and Compose for more information. |
Next to "Unsharp Mask" in the Filters menu is another filter called Sharpen, which does similar things. It is a little easier to use but not nearly as effective: our recommendation is that you ignore it and go straight to Unsharp Mask.
In some situations, you may be able to get useful results by selectively sharpening specific parts of an image using the Blur or Sharpen tool from the Toolbox, in "Sharpen" mode. This allows you to increase the sharpness in areas by painting over them with any paintbrush. You should be restrained about this, though, or the results will not look very natural: sharpening increases the apparent sharpness of edges in the image, but also amplifies noise.
When you take pictures in low-light conditions or with a very fast exposure time, the camera does not get enough data to make good estimates of the true color at each pixel, and consequently the resulting image looks grainy. You can ΓÇ£smooth outΓÇ¥ the graininess by blurring the image, but then you will also lose sharpness. There are a couple of approaches that may give better results. Probably the best, if the graininess is not too bad, is to use the filter called Selective Blur, setting the blurring radius to 1 or 2 pixels. The other approach is to use the Despeckle filter. This has a nice preview, so you can play with the settings and try to find some that give good results. When graininess is really bad, though, it is often very difficult to fix by anything except heroic measures (i.e., retouching with paint tools).
Every so often you have the opposite problem: an image is too crisp. The solution is to blur it a bit: fortunately blurring an image is much easier than sharpening it. Since you probably don't want to blur it very much, the simplest method is to use the ΓÇ£BlurΓÇ¥ plug-in, accessed via Filters->Blur->Blur from the image menu. This will soften the focus of the image a little bit. If you want more softening, just repeat until you get the result you desire.
There are two kinds of objects you might want to remove from an image: first, artifacts caused by junk such as dust or hair on the lens; second, things that were really present but impair the quality of the image, such as a telephone wire running across the edge of a beautiful mountain landscape.
A good tool for removing dust and other types of lens grunge is the Despeckle filter, accessed as Filters->Enhance->Despeckle from the image menu. Very important: to use this filter effectively, you must begin by making a small selection containing the artifact and a small area around it. The selection must be small enough so that the artifact pixels are statistically distinguishable from the other pixels inside the selection. If you try to run despeckle on the whole image, you will hardly ever get anything useful. Once you have created a reasonable selection, activate Despeckle, and watch the preview as you adjust the parameters. If you are lucky, you will be able to find a setting that removes the junk while minimally affecting the area around it. The more the junk stands out from the area around it, the better your results are likely to be. If it isn't working for you, it might be worthwhile to cancel the filter, create a different selection, and then try again.
If you have more than one artifact in the image, it is necessary to use Despeckle on each individually.
The most useful method for removing unwanted ΓÇ£clutterΓÇ¥ from an image is the Clone tool, which allows you to paint over one part of an image using pixel data taken from another part (or even from a different image). The trick to using the clone tool effectively is to be able to find a different part of the image that can be used to ΓÇ£copy overΓÇ¥ the unwanted part: if the area surrounding the unwanted object is very different from the rest of the image, you won't have much luck. For example, if you have a lovely beach scene, with a nasty human walking across the beach who you would like to teleport away, you will probably be able to find an empty part of the beach that looks similar to the part he is walking across, and use it to clone over him. It is quite astonishing how natural the results can look when this technique works well.
Consult the Clone Tool Help for more detailed instructions. Cloning is as much an art as a science, and the more you practice at it, the better you will get. At first it may seem impossible to produce anything except ugly blotches, but persistence will pay off.
In some cases you may be able to get good results by simply cutting out the offending object from the image, and then using a plug-in called ΓÇ£ResynthesizerΓÇ¥ to fill in the void. This plug-in is not included with the main GIMP distribution, but it can be obtained from the author's Resynthesizer web site. As with many things, your mileage may vary.
When you take a flash picture of somebody who is looking directly toward the camera, the iris of the eye can bounce the light of the flash back toward the camera in such a way as to make the eye appear bright red: this effect is called ΓÇ£red eyeΓÇ¥, and looks very bizarre. Many modern cameras have special flash modes that minimize red-eye, but they only work if you use them, and even then they don't always work perfectly. Interestingly, the same effect occurs with animals, but the eyes may show up as other colors, such as green.
GIMP does not include a special tool for removing red-eye, but it isn't all that hard to do. Basically the idea is to zoom the area around the eye so that it is nice and large and easy to work with; then make a selection of the red part of the eye and a bit of the area around it; feather the selection so that you don't create sharp-looking edges; and finally desaturate the red channel inside the selection using one of the color tools---Levels, Curves, or Hue/Saturation. It takes a little practice the first few times, but once you have the technique mastered, you should be able to quickly and easily create quite a natural looking eye color.
If you would like a more automated approach, you can try downloading a recently created redeye plug-in from the GIMP Plug-in Registry. We have not received any feedback so far about how well it works. It comes in source code form, so you will need to be able compile it in order to use it. (See Installing New Plug-ins for information on how to do this.)
What file format should you use to save the results of your work, and should you resize it? The answers depend on what you intend to use the image for.
If you intend to open the image in GIMP again for further work, you should save it in GIMP's native XCF format (i. e., name it something.xcf), because this is the only format that guarantees that none of the information in the image is lost.
If you intend to print the image on paper, you should avoid shrinking the image, except by cropping it. The reason is that printers are capable of achieving much higher dot resolutions than video monitors---600 to 1400 dots per inch for typical printers, as compared to 72 to 100 dots per inch for monitors. A 3000 x 5000 image looks huge on a monitor, but it only comes to about 5 inches by 8 inches on paper at 600 dpi. There is usually no good reason to expand the image either: you can't increase the true resolution that way, and it can always be scaled up at the time it is printed. As for the file format, it will usually be fine to use JPEG at a quality level of 75 to 85. In rare cases, where there are large swaths of nearly uniform color, you may need to set the quality level even higher or use a lossless format such as TIFF instead.
If you intend to display the image on screen or project it with a video projector, bear in mind that the highest screen resolution for most commonly available systems is 1600 x 1200, so there is nothing to gain by keeping the image larger than that. For this purpose, the JPEG format is almost always a good choice.
If you want to put the image on a web page or send it by email, it is a good idea to make every effort to keep the file size as small as possible. First, scale the image down to the smallest size that makes it possible to see the relevant details (bear in mind that other people may be using different sized monitors and/or different monitor resolution settings). Second, save the image as a JPEG file. In the JPEG save dialog, check the option to ΓÇ£Preview in image windowΓÇ¥ , and then adjust the Quality slider to the lowest level that gives you acceptable image quality. (You will see in the image the effects of each change.) Make sure that the image is zoomed at 1:1 while you do this, so you are not misled by the effects of zooming.
See the File Formats section for more information.
As in most softwares, in GIMP, printing needs to go to main menu
→ . However it is very usefull to have in your mind some elementary concepts to avoid you unpleasant surprises when looking at the result or to cure it if that occurs. You already must remember:that image displayed on the screen is in RGB mode and printing will be in CMYK mode; consequently color feature you'll get on printed sheet will not be exactly what you was waiting for. That depends on the used corresponding chart. For the curious ones some adding explanations can be got through a click on these usefull Wikipedia links:
that a screen resolution is roughly within a range from 75 up to 100 dpi; a printer resolution is about 10x higher (or more) than a screen one; printed image size depends on avalaible pixels and resolution; so actual printed size doesn't correspond inevitably to what is displayed on screen nor avalaible sheet size.
Consequently, before any printing it is relevant to go to: print sizeΓÇ¥ box adjusting either sizes or resolution; you can see that the both values are linked without possibility to dissociate them. You can see too that you can dissociate x and y resolution but it is risky! Probably this possibility is open because printers are built with different x vs. y resolutions. Nevertheless if you unlinked them you can be very surprised! You can try this in special effects.
→ and choose here your convenient output size in “Last recommandation: think of checking your margins as well as centering. In this case modify them and go into the menu
→ . It would be a pity if a too much large margin cuts off some part of your image or if an inapropriate centering damages your work espacially if you use a special photo paper.Modern digital cameras, when you take a picture, add information to the data file about the camera settings and the circumstances under which the picture was taken. This data is included in JPEG or TIFF files in a structured format called EXIF. For JPEG files, GIMP is capable of maintaining EXIF data, if it is built appropriately: it depends on a library called “libexif”, which may not be available on all systems. If GIMP is built with EXIF support enabled, then loading a JPEG file with EXIF data, and resaving the resulting image in JPEG format, will cause the EXIF data to be preserved unchanged. This is not, strictly speaking, the right way for an image editor to handle EXIF data, but it is better than simply removing it, which is what earlier versions of GIMP did.
If you would like to see the contents of the EXIF data, you can download from the registry an Exif Browser plug-in . If you are able to build and install it on your system, you can access it as Filters->Generic->Exif Browser from the image menu. (See Installing New Plug-ins for help.)