masks on darktable 5.6 user manual

overview

Mon, 01 Jan 0001 00:00:00 +0000

Masks allow you to limit the effect of a module so that it only applies to certain parts of the image.

A mask can be regarded as a grayscale image where each pixel has a value between 0 and 1.0 (or between 0% and 100%). This value is the opacity and is used to determine how much a module affects each pixel.

The following sections explain how to construct masks in darktable.

drawn masks

Mon, 01 Jan 0001 00:00:00 +0000

With the drawn mask feature you can construct a mask by drawing shapes directly onto the image canvas. Shapes can be used alone or in combination. Once a shape has been drawn on an image it can be adjusted, removed, or reused in other modules.

Shapes are stored internally as vectors and are rendered with the required resolution during pixelpipe processing. Shapes are expressed in the coordinate system of the original image and are transformed along with the rest of the image by any active distorting modules in the pipe (lens correction, rotate and perspective for example). This means that a shape will always work on the same image area regardless of any modifications that may be subsequently applied.

parametric masks

Mon, 01 Jan 0001 00:00:00 +0000

The parametric mask feature offers fine-grained selective control over how individual pixels are masked. It does this by automatically generating an intermediate blend mask from user-defined parameters. These parameters are color coordinates rather than the geometrical coordinates used in drawn masks.

For each data channel of a module (e.g. Lab, RGB) and several virtual data channels (e.g. hue, saturation) you can construct a per-channel opacity function. Depending on each pixel’s value for a given data channel this function calculates a blending factor between 0 and 1 (100%) for that pixel.

combining drawn & parametric masks

Mon, 01 Jan 0001 00:00:00 +0000

Drawn and parametric masks can be used in combination to form a single mask that can be applied to a module.

There are two main elements which control how individual masks are combined: the polarity setting of each individual mask (defined by the plus or minus buttons) and the setting in the “combine masks” combobox.

The “combine masks” combobox contains the following options, defining how the drawn and parametric masks will be combined:

mask refinement & additional controls

Mon, 01 Jan 0001 00:00:00 +0000

When a mask is active, several additional sliders are shown which allow the mask to be further refined.

details threshold: This control allows you to alter the opacity of the mask based on the amount of detail in the image. Use this slider to select either areas with lots of detail (positive values) or areas that are flat and lacking in detail (negative values). The default (zero) effectively bypasses details refinement. This is mostly useful to apply sharpening and blurring effects that ignore out-of-focus (bokeh) regions or to sharpen only blurry parts, preventing over-sharpening of in-focus regions.

Note: The data used for the detail mask refinement is taken from the demosaic or rawprepare stage in the processing pipeline, not from the module’s input (which is used for the other parametric mask criteria). None of the processing modules after demosaic will have any effect on the detail mask and it is not available for non-RAW images.

AI masking

Mon, 01 Jan 0001 00:00:00 +0000

AI masking requires AI features to be enabled in AI preferences, with a model active for the AI task the specific masking tool uses (each tool below states which one). See the AI features overview for what runs where.

Currently darktable provides one AI masking tool:

AI object: Uses the mask task. Click on a subject to produce a mask around it; iterate by adding more clicks until it covers exactly what you want. The result is vectorised into a group of path shapes and can be edited like any other drawn shape – see drawn masks: AI object for details.

raster masks

Mon, 01 Jan 0001 00:00:00 +0000

As described in the previous sections, the final output of a module’s mask (the combined effect of any drawn and parametric masks) is a grayscale raster image representing the extent to which the module’s effect should be applied to each pixel. This raster image is stored internally for active modules and can be subsequently reused by other modules in the pixelpipe.

As with any mask, if the opacity value for a pixel in a raster mask is zero, the module’s input is passed through the module unchanged. If the opacity is 1.0 the module has its full effect. For each value between 0 and 1.0 the module’s effect is applied proportionally at that location.