Seamless Cloning

The third new tool, Seamless Cloning, has the potential to revolutionize cloning in Gimp because of its many properties: Gimp achieves the "seamless" effect by automatically adjusting the edge between the inserted object and the target background. This works much like the Heal tool but does not require manual adjustments. Additionally, seamless cloning makes it easier to crop motifs, which you no longer need to compose exactly to allow seamless insertion.

To use Seamless Cloning, first open the background image into which you want to paste the image and then open the image you want to insert. You can and should be generous in selecting the second image: If you define the object you want to insert too precisely, Seamless Cloning will not have a sufficiently large edge area with which to achieve a genuinely seamless effect.

Next, copy the image you want to insert into the target image and convert the floating selection into a normal layer, which you then make invisible. Now, enable the background layer into which you want to insert copied material. Make sure at this point that no other selection exists in the image. Now enable the Seamless Cloning tool, currently characterized by the same symbol as the Move tool (the four-direction "compass").

After clicking on the background layer, Gimp starts to convert the material you want to insert. It could take some time before the results appear. You can still move the material that this step creates by dragging it with the mouse; Gimp will automatically adjust the brightness of the inserted material to match the background (Figure 8). Pressing the Enter key or enabling another tool completes the paste action.

Figure 8: (Top left) The cropped original image and (bottom left) the version used for Seamless Cloning. This image was inserted twice into the background image (right). Note how well Gimp has adjusted the edges.

At this time, the tool works very slowly and – especially with large images – isn't particularly stable. Additionally, Seamless Cloning inserts the cloned image material directly into the background. If it were to use a separate layer, you could later mirror or scale the results. For this reason, you should always paste into a copy of the background layer, which you can then continue to edit.

The developer of this tool describes this process in his blog [7]. A longer paper describes the procedure on which the tool is based [8] and explains what exactly you can do with it.

Great Color Depth

Gimp 2.9 already optionally works with the higher color depth intended for Gimp 2.10; in other words, it can edit images with more than 8 bits per RGB channel. This means you can take images that were created with a RAW converter and that have a color depth of 16 bits, load them directly into Gimp without conversion loss, and proceed to edit them.

Gimp 2.9 loads images with this color depth in the normal way and automatically enables the matching functions. Alternatively, you can set the desired color depth after loading. The features required to do this are available under Image | Precision (Figure 9).

Figure 9: Gimp version 2.9 supports greater color depths, which you can also configure manually.

With a color depth of 16 bits, you can generate real HDR. Moreover, greater color depths give you a reserve that preserves the finer details in highlights and shadows and lets you apply more complex functions, such as denoising or sharpening. Visible artifacts produced by functions operating on images with 8 bits of color depth per RGB channel can thus be eliminated in images with 16-bit color depth (see also the "Color Spaces" box).

Non-Destructive Editing

The use of GEGL in all parts of the image editing program offers another decisive advantage – that is, non-destructive image editing. You will be familiar with this principle from RAW converters, which initially open a RAW image in read-only mode when editing and then convert the image into a bitmap. This leaves the original image completely unchanged. All other actions the RAW converter then provides are applied to the bitmap copy of the image in memory.

The software records all actions in the form of recipes and then applies them to the image data. This process makes it easy to restore the original image from a copy of the image data. Additionally, you can change recipes at any time and re-apply them to the image data, thus creating any number of image variants.

This process is also referred to as graph-based editing, which allows you to change the recipe or create a branch at any point. For example, you could adjust the brightness then white balance, then crop the image; you could first correct white balance then brightness, and finally crop; or you could first crop the image then perform the other actions.

Interestingly, you will get different results with all three variants, because the initial data for following steps are modified slightly by preceding steps. However, this doesn't necessarily have to be true: If you use strictly structured programs, such as Darktable, which was designed for maximum reproducibility of results, the end result will be totally different from, for example, RawTherapee [10], which is oriented to a far greater extent on the workflow of previous Gimp versions.

In Gimp 2.10, this procedure will revolutionize image editing [11], because you can retroactively add editing steps that you forgot to perform to the recipe. In this way, you can branch off an established workflow, with the option of testing variants, without having to keep innumerable temporary files on your hard disk [12].

You can also easily implement a macro recorder in this way: All you need to do is record, manage, and display your recipes, instead of the functions or even the resulting bitmaps.