Due to the standardization of sRGB on the Internet, most image formats use SRGB as the default working color space. If the color space of an image is unknown and it is an 8- to 16-bit image format, assuming it is in the sRGB color space is a safe choice. This extends to grayscale as well. We assume non-linear grayscale. These assumptions are overridden if a particular image format includes color space and / or gamma metadata. ImageMagick assumes linear color if the color space is RGB instead of sRGB. You can also override the default color space assumptions with color profiles or the -set option.

ImageMagick supports color profiles, however, for images without a profile or a declaration of colorspace, ImageMagick assumes non-linear sRGB. Most image processing algorithms assume a linear colorspace, therefore it might be prudent to convert to linear color or remove the gamma function before certain image processing algorithms are applied. For example,

convert myimage.jpg -colorspace RGB -resize 200% -colorspace sRGB mybigimage.jpg

To declare that an image is linear RGB rather than sRGB, you can use the set option:

convert myimage.png -set colorspace RGB myRGBimage.png

Note that declaring an image as linear is not the same as converting the image to linear. Declaring it is linear only sets the meta data and does not change the pixel data. Whereas converting to linear actually changes the pixel data as described in more detail below.

Afterwards, the verbose information for the output file lists the colorspace as RGB. This only works on image types containing meta data that distinguishes between linear RGB and non-linear sRGB, such as PNG and GIF. Therefore, if the above command is run with a JPG or TIF output format, the verbose information for the colorspace still shows sRGB. In order to properly have the JPG output know that it is linear RGB, include an appropriate color profile.

By default, converting color images to grayscale returns non-linear gray. To instead convert to linear gray, use the -set or -intensity options:

convert myimage.png -set colorspace RGB -colorspace gray myRGBimage.png
convert myimage.png -intensity Rec709luminance -colorspace gray myRGBimage.png

The same concept is needed when separating channels. Normally, the conversion to separate each channel of an sRGB color image produces separate non-linear grayscale images. However the same concept can be applied, if it is desired to keep the separate channels as linear grayscale. For example, the following produces linear grayscale channels.

convert myimage.png -colorspace RGB -separate myimage_channels_%d.png

When separating and recombining channels, with potential intermediate processing, it is important to identify the colorspace used, especially during the recombination. For example,

convert myimage.png -separate myimage_channels_%d.png
convert myimage_channels_*.png -combine myimage2.png

In the above example, the separated channels are non-linear and the default for -combine is to assume non-linear channels. So the result is the same as the input.

If one desires to separate to linear grayscale channels, recombine them later back to non-linear color, perhaps after some processing, then use the same concept as above for maintaining linear grayscale:

convert myimage.png -set colorspace RGB -separate myimage_channels_%d.png
convert myimage_channels_*.png -combine -colorspace RGB -set colorspace sRGB myimage2.png

When converting to another colorspace and back, such as between sRGB and HSL, the following two commands handle the first case of linear channels and the second case of non-linear channels:

convert myimage.png -colorspace HSL -separate myimage_channels_%d.png
convert myimage_channels_*.png -set colorspace HSL -combine -colorspace sRGB myimage2.png
convert myimage.png -set colorspace RGB -colorspace HSL -separate myimage_channels_%d.png
convert myimage_channels_*.png -set colorspace HSL -combine -colorspace RGB -set colorspace sRGB myimage2.png

A majority of the image formats assume an sRGB colorspace (e.g. JPEG, PNG, etc.). A few support only linear RGB (e.g. EXR, DPX, CIN, HDR) or only linear GRAY (e.g. PGM). A few formats support CMYK. For example JPG does, but PNG does not. Then there is the occasional format that also supports LAB (that is CieLAB) (e.g. TIFF, PSD, JPG, JP2). For additional information, see the Colorspace and Supported Formats pages.

When specifying individual colors as rgb(...) or hex, these colors will still be interpreted as non-linear, that is, as sRGB colors. However if one wants to create linear colors, use icc-color(rgb,r,g,b)", where r, g, and b are in the range 0 to 1. See the Color page.

There are now also distinctions between linear and non-linear shades of gray. Any named shade of gray, such as black, white or numbered grays like gray50 are non-linear gray. However, gray(...) is now linear gray. Hex grayscale values are non-linear.

This means that you need to be careful when you create grayscale gradients. For example, gradient:, gradient:"white-black", gradient:"gray100-gray0" generates non-linear gradients, however gradient:"gray(255)-gray(0)" or gradient:"gray(100%)-gray(0%)" generates linear gradients.